Začátek vypadá skvěle a veze se na vlně AI, které k nám vtrhlo jak velká voda a vypadá, že s námi zůstane. Za mě naštěstí už na konci první stránky – byť první kapitola je vlastně celá o ní – se AI vytrácí a zůstávají strategické návyky pro lepší rozhodnutí – všechny ty věci, o kterých Honza posílá newslettery, já je sleduji zpovzdálí a zatím jsem moc nechápal. Tady jsou skvěle shrnuté a to je pro mě mega přidaná hodnota.

Strategické cykly a rozhodnutí (test, abyste si řekli opak strategického rozhodnutí a pokud je to zjevně hloupost, tak nejde o strategické rozhodnutí, mi přijde fantastický a jednoduchý), tři délky cyklů pro které je třeba plánovat, rozdíl mezi plánováním z budoucnosti a adaptací z přítomnosti.

Nastavování cílů a v jakých hladinách je plánovat – on je totiž rozdíl mezi výstupy, výsledky a dopady. Výstup zvládne každý, u výsledků musíte přemýšlet jak se k nim dostat, dopady už je mega výzva.

Zákaznické výzkumy. Opravdu je děláte nebo si jenom povídáte s lidmi?

Warleyho mapy Honza používá rád, nikdy jsem je nepochopil, až teď. Kecám, asi jsem je pořád plně nepochopil, ale konečně mi začaly dávat smysl, důležitost horizontální osy jsem totiž neznal a ta dělá celý ten rozdíl!

Principy budování organizace, kterých je „směšných“ 40 a jsou rozdělené do 4 fází – nemá smysl zavádět ty, na které nejste připravení. Jednoduchá aplikace vám je pomůže zmapovat a pak vzhůru na ně!

Obrana před konkurencí, řešení (a hledání) toho důležitého, rozhodování podle situace (ano, podle typu situace byste se měli rozhodovat jinak), ověřování nápadů a kdy si to zjednodušit a kdy zkomplikovat a to celé s praktickými příklady jak to dělají v House of Řezáč případně na vymyšlené palačinkárně. Za mě skvělá kniha na otevření očí, na další nápady, na věci, které hrají spolu dohromady a mohou váš byznys někam posunout. Spousta odkazů na další zdroje, takže přečtení knihy samotné není konec, ale opravdu jenom začátek cesty. Vydejte se na ní!

Předobjednat si knihu můžete na jeho webu.

Zápisky:

• stačí si o věcích číst nebo je musíte zažít?
• rituály, dýchání, výzvy
• sdílení, kde věci cítíte a odkud o nich mluvíte a do jaké hloubky jdete
• rozvoj není konzumace obsahu, ale prožívání a trávení věcí
• můžeš zapomenout? Nebo musíš ty situace vyřešit a doprožít?
• rychlost nebo pomalost? Lepší je mít čas věci nechat uležet a prožít je
• stačí být slyšen, ne dostávat rady
• jaký máte vztah s rodiči? A jak moc vás to ovlivňuje i když si třeba říkáte že ne?
• seriál Já jinak, pokud chcete přemýšlet o něčem co nemůže fungovat ale dost možná funguje
• vision quest – 4 dny sám v lese bez jídla, jak moc vás to může změnit a kam vás to posune?
• přechodové rituály jako ztracený nástroj společnosti – vědomý přechod z jedné životní fáze do druhé. Cesta na zkušenou bývala úplně normální
• muži emoce mají – jen se nenaučili o nich mluvit
• Je potřeba žít život, nechodit jenom na semináře – integrace do každodenního života je důležitější než hromadění zážitků z výcviků

Co dál je otázka, která mi zůstává. Marcel je na tom lépe, ví že chce pokračovat se Síla duše, já částečně doufám, že Jozef vymyslí pokračování a do toho se zkouším rozhlížet kolem, jak si tyhle věci udržet, kým se obklopit.

But now I found the time to check every single presentation (they are all on our website) from the day and learnt the following. Actually before I did my manual work I asked AI (Google NotebookLM specifically) to summarize it for me – and I learnt nothing from that output, obviously I’m still person who needs to invest own time and energy to have a feeling of achievement.

Intro keynote about DevOps was great – while I introed John with a story where I wasn’t 100% confident about inviting him, it worked out greatly. 820 metadata types deployable via Metadata API, every team has to do something to allow their metadata to be deployed (and some done do so, but they are constantly try to close the gap). We heard some stories from the background and learnt about StupidApexWorkAround class, about object Id which has to be changed just before going to production, about migration from managed package to platform native solution, which has been built in 6 months (to Pilot) thanks to AI help. Take-away: DevOps 2.0 is completely different product, I need to check it again.

Marketers

Olga and Timo made me excited about Marketing Cloud Next, additional channels and the whole platform unification topic. I would argue about respectful timing on Saturday at 10am, but I might happily ignore that WhatsApp message without being too annoyed. But scaling by using WhatsApp, automated qualification, RAG search (it is here again!) looks too good to be true. Lessons learnt and implementation reality is even more surprising and humbling – start small, one tweak at a time, don’t panic from first tests (takes a while to settle down), 4 – 6 weeks for basic implementation (including testing and iteration!).

Hanna and her „small“ tips for future success, valid not only for Marketing Cloud – naming conventions, description fields, monitoring, data diet (have only what you need, not 500 fields just in case), remove what is not needed anymore (and also have a plan from beginning when you can remove every single thing) and least privilige approach. With 2 things to do this week if you don’t have time. Awesome!

David’s talk was most likely a nice add-on to Hanna – a lot of great things how to look at what the system is doing, how to spot journeys which doesn’t run as expected (well, that actually means you can define the KPIs) and all the other monitorings you can do in Marketing Cloud. Not much to gain based on the slide to someone who don’t know (me) as a marketer definitely eye opening to what I should be doing.

Eva and emailing in scale – surprisingly it wasn’t that much about the technical part (how to avoid being marked as spam) but rather the organization behind – with 2M monthly emails and 2000 journey across 100 users I’m not surprised. One email template with referenced content blocks to your survival, integration with Figma if just by using the same name for relevant parts and … approvals!

Vicki covered Data360 nicely, but when I saw the picture above I had to wink – why the hell? But following through the presentation it probably makes sense, Data360 gives you the possibility to calculate additional metrics, segmentation can be driven by clicks not code (I hated this part of MCE), more types of segments, and much more. Obivously – especially in bigger orgs – you will need to understand all the technologies and its impact on your bill – when to sync, when to leverage what just to save your credits. Or maybe we won’t even worry about it anymore and just pay.

David followed with security in Data360 – maybe not the low level security as it isn’t really a storage, rather data in motion. More about the „perceived“ security we just talk at one customer as well. When you over ingest (we already mentioned it above) because it feels it makes sense rather than later extend the sync, when you segment (ideally super cleverly) and this segmentation reveals data to which the people would not originally have access/thought about and finally you activate/push the data to other system and suddenly everyone can see this indication, which was originally just a small internal label but now it says something very specific about the profile. And that’s the moment – way way downstream – where you don’t want to get, but because a lot of things have to happen together you cannot really foresee it

Ekat and Marketing Cloud on Core, should we care? It is actually scary as admins are getting more and more work it looks like with this they will gain even more – security, shared data with marketing (kind of 100% shared), Data360, flows instead of journey builder (so admin can do it, right, no need to pay marketer). So while I see the career expansion opportunity I’m scared for them as well. Luckily (?) MCE is here to stay at the moment, so let’s see how quick the transition will be.

Filip and BlockSDK with AI – no idea, this topic is over my head, but there is a link to public app to play with that!

AI

Robert – the most handsome speaker of CzechDreamin (official title) – generated his presentation with AI and had a full room of excited people. He shared My Org Butler, a tool which showcase what is possible. And the presentation is full of tips about less if often more, about finding balance, hybrid search which beats SOQL, how RAG on documents helps with finding answers, that Data Cloud Foundations is free and we should start now because it will be just more complex later, that LLM are trained on old data but fresh data are just one API call away, how persistent memory is crucial, or how headless is the future where we want to go. Take-away: try the My Org Butler

Crawl, walk, run approach for AI adoption by Yosra and Jim, the need to start simple (again this approach!), you the right technology for the right task, have clear business requirements and quality data, concise instructions which aren’t over complicated/constaining (reliability dropped past 8 intructions in production for Yosra), 20+ iteration cycles on master prompt as small wording can shift the behaviour – this would kill me! Edge cases and intents you didn’t foresee to come and need to constantly monitor for. Take-away: deploy AI agents will not be quick job and prompt writing is obviously a still key job.

Pei and her AI says yes but … thinking about it one more time AI is so much similar to us – it doesn’t know what it doesn’t know, it has been trained on patterns not context (every client is different but they are all the same), and it can sound very authoritative (it is correct because it is nicely looking). It reminded me so much about job in consulting And the mirror you should look at it is crucial – what assumptions you made, do you have honest answer from where the things came, can you defend every element, what might be wrong/incomplete with your answers, is it easy to maintain (otherwise it is technical debt even though you are so proud about it), use client’s terminology.

Lillie and her case for AI in Experience Cloud – don’t throw it on top of your problems, it will only amplify them. Deflection is earned by good design, not bought with an agent. If you have just unstructured data the AI-generated summaries won’t be the best, have good structured knowledgebase first. If you personalize on iferred attributes it is just creepy or presumptive, get your data correct first. Don’t deploy AI on emotionally-loaded topics (billing disputes, compaints), the empathy should stay with people.

Developers

Nicolas covered sfdx-hardis which looks better and better every time I see it. And while it does basically what a good CI/CD pipeline does, this is – in my view – focused more on the regular admin who don’t really understand those things. Also covered SFDMU and External Connected App – one reason I didn’t pay much effort into their migration so far is, that I didn’t know that all the credentials will stay the same! Take-away: look deeper into sfdx-hardis and really use it on a project. The documentation it generated on one of them was really good.

Igor about mega-orgs – actually when I heard at ArchitectDreamin about an org with 250 millions of contacts, that’s something I can hardly imagine (because I just don’t know country with so many people). Anyway, he walked us through the key limits (30 millions of ContentDocuments only, really?), how to use Grafana for DYI monitoring and costs optimization. Take-away: on top of Nebula Logger and sfdx-hardis try to find time to look at Grafana as well.

Michal and LWC architecture – thing I never felt good enough. While going through the presentation it looks simpler now (and way harder at the same time) – clear boundaries need to be set. The presentation contains a link to LWC toolkit with a decision tree to get you on the best path for your project.

Keir and Prompt Templates, because they aren’t just for agents and can be used directly from Apex. Why? Because Apex is not great at getting meaning, insights or context from text or in generating text with the right tone and language, which summarize some key data. I’ve never really thought about it but it can be super powerful and the examples are really simple code to achieve big things.

Ever saw a bad code? Houssam and Fátima claims that 99% of junior developers didn’t, which surprised me. Seeing the slides I would love to run a pannel with them and Piotr next time as they have opposite opinions (hard-code messages or not?), but besides that a lot of known anti-patterns (DML in loops), SOQL injections, Platform Cache (never saw it live), switch and „when else“ clause which should be always added (no clue why, but they probably said it), triggers and much more. I have to wonder how they define „junior“ because half of the presentation feel even to me like the absolut norm aligned with the best practices how to survive on the platform.

Todd to whom Lukas would be willing to listen even if speaking about toilet paper (actually a few people picked on introduction Lukas did of them in his video). Well, I didn’t know about the document.designMode = „on“ which allows you just simply type on any webpage, always did it via the Console. Tons of videos in the presentation so you’ll miss the talk but not the action. How to make the console log look better. That you don’t need to deploy a component to change it for testing. Take-away: rewatch this presentation!

I saw Michal’s presentation about developer who wasn’t there at Prague User Group and wanted it here as well. Demo is the main part, so you won’t find much in the presentation plus the presentation is probably too heavy even for regular developer. Based on that I realized how much the train left me and how AI will really change the delivery world. We even recorded a podcast episode about this (in Czech, sorry).

Beech had a super heavy presentation and while it is hard just based on the presentation restore the whole speach, some elements are clear – do NOT store secrets in your code repository, there are better ways how to inject them when needed. Anypoint Platform API Visualizer looks great, think twice where you can survive with Fire & Forget pattern, OpenAPI approach, DataWeave – still didn’t find energy to play with that.

Business & Admins

Marin and his being Agile – first of all I have no clue how he could deliver 118 slides in allocated time, but they are just great, similar style to Pei. If you are already scared from opening it don’t be, it is very good comics which reminds you that being Agile doesn’t mean daily stadups, reviews, tracking velocity, etc. but the 12 principles behind the manifesto are pretty much different. And if 12 is too much you should start with 4 – make sure that it satisfy the customer (understand personas, business model and value proposition), T-profile people might be better and working product is the crucial one – and if it isn’t delivered it doesn’t exist.

Aaron and fault path was mostly demo, but from those a few slides – store the errors in custom object (Nebula Logger?), create dashboard, think about retention policy and open all the relevant questions – do we need more user training, is that task too complex for flow, are we even using the right flow type?

I just love Lukasz intro picture. And sadly I saw automation hell where one change triggers something completely else which will …. Governor limits which politely inform you that your design sucks and should not be overcome by going asyns. The graveyard of old fields on overloaded object and finally that sharing is architecture as it drives performance and everything. Must have been great talk!

Maria and Justyna and their take on organizational changes – I never knew there are so many frameworks to drive the change, obviously discipline on its own. Also after seeing that – I’m not surprised that majority of projects fail in it, because people rarely see the complexity behind.

Matt’s session was another often mentioned, with the AI around the average time-to-exploit is down to 20 hours (and if that feels ok, you should know it used to be 2,3 years), so it is just a matter of time when it happens to you and you should be ready to act. Just open the presentation it has all the steps you should take and questions to ask. Golden!

Jasmine and decisions which will cost you later nicely add to Hanna, Lukasz and others, the presentation includes 5 (easy) things to do now. But we are back to proper object modelling (and no-one said the SF one is the best for everyone), automation sprawl and order of execution (that surprise so many people), hard coded IDs (well, wait for Piotr), naming conventions and governor limits. Easy to scroll through but it might open so many questions in your head.

Milan’s presentation about flows open a lot to imagination, would be definitely worth hearing the talk. But even from those screenshots you fantasy can start with what might be possible and how it might improve use experience. And then you go the route to DRY principle and how to tackle the actual flow creation and composition. Really nice one.

Paul and Roger and ADHD – not a standing room as at London’s Calling but still a handful of people attended and I realized why I should not trust the event coordinator that anyone can manage the room without microphone. A lot of excercises but to me the most crucial slide of the presentation is Roger’s „about me slide“ (#24) – instead of all the general info no-one really cares about it lists what works for him and why he migh behave during the meeting differently than you would expect. That’s something I should put together as well.

Rich and great roadmaps – there was probably way to many great things hidden in the talk, which I now miss Fantastic pictures so potentially a really funny presentation which I need to watch somewhere. A process how to make roadmap, reasons why to make it. Tools which will help.

So much of the value in a roadmap IS the process!

I simply love Andre’s presentation about data quality being a leadership problem. They typically don’t want to hear it, but if data is a strategic assets they should treat it that way, right? Governance should not be about more documents, admin work or meetings, but about clarity – who owns, what is the standard, which fields are the most crucial and who checks them. Salesforce admin should be a facilitator not janitor – spot patterns, make issues visible, build dashboards and advice on controls. The managers should explain why data is important and ensure the team follows the process. They also needs to know who creates the data and which decision depends on them and what happen wheny they are wrong.

Leanne and Liz, another first-time presenters at CzechDreamin (actually we had about 50% of speakers here for the first time, really suprised us such number) spoke about surviving global rollout. The feedback from audience was mainly about great team play during the presentation and the presentation is full of golden nuggets – weak governance, last minute changes, too many stakeholders, treating UAT as theatre (been there, saw that). Repeated what we heard during the day already – small bets, smart adjustments, why we are doing it? If something is straighforward it is usually a warning sign. That technology is rarely the hardest part, humans are! The closer to go-live the smaller decision group should with clear ownership. Work with timezones, don’t fight them. Commit to one system! Involve users early. And that hypercare is the moment where trust is won or lost. This was awesome even on slides only!

Melissa spoke about tracking ROI of your automation and that it should be just about financial savings and ignoring maintenance & training. You should also look for qualitative gains, user satisfaction, hours of work saved. Custom object, some simple triggers to fill it with data (lead assignment time, response time, conversion rate, etc) and then you can report also how many new deals you got and everything. I just never did the measurement before as typically my customers don’t know them, so never got a great looking numbers how we improved the system

Chandan spoke about the new UI improvements in flow, how we should choose the smallest pattern that solves the UX problem and then reuse it with Custom Labels everywhere. That these days there might not be a need for custom LWC as flow with a bit of styling, sections and thought can be enough.

Good old times when clients were happy with the standard UI are gone.

I saw Louise presenting on User Access Policies at IrishDreamin and while I never implemented it yet (either have clients to small to benefit from it, I know, wrong feeling) or someone else was responsible for it, it really feels like a powerful feature, which is surprisingly included for free everywhere. Such a time saver to an admin after implementation as it will automatically (de-)provision the right licenses, permission set (groups) and much more. All of that included with a high-level view how to approach the whole topic of profiles, permission sets and permission sets groups.

Consultants

I actually sit in Kedar’s session for a while as I’m constantly thinking whether to dig deeper into OmniScript or don’t bother anymore as it will be going away (I heard, but who knows). It is really interesting how similar it is to flows and while we play a few scenarios to make the best decision at the end it was (and probably is in reality as well) pretty schematic. OmniScript used to be better on UI but with the latest advancements of flow probably not that much different anymore. API calls, data raptors and all these things are potentially good enough in flows as well and what stay for OmniScript is the high effort of customization – yes, configure the thing was always hard and need a lot of knowledge and clicks so the final takeaway – start with flow by default is what stayed with me (as there is no universal winner).

CPQ or Agentforce Revenue Management by Oleg & George is fabulous. People complain about this new product not being on parity so how could Salesforce stop selling CPQ, but just going through the presentation I just like ARM much more. More end-to-end process support (not just quoting), product catalog which probably makes more sense (as I can sell one product in multiple ways, there is no reason to clone it), transaction editor easily extendable with flows (everything is flow basically in ARM), pricing engine which feels so easy to read compared to the old mix of Pricing Rules, Summary Variables and everything else, document generation which can be used for anything not just quotes. Yes, the migration is not lift and shift, that would kill you quickly, but with the right approach it feels like a product which have been thought through.

Speaking about products we typically don’t cover at CzechDreamin, Prag came to Prague with his Field Service talk. I like how he covers the non-typical capabilities, what is not covered in the presentation is sadly all the gotchas – crews look great but you need to constantly define them, complex work is awesome but has impact in scheduling and there is not templating for that. Maybe it has been said, maybe not, still great to have one more contact in the ecosystem with whom I can share my experience and ask for help.

Slack, had we ever have it before? Piyusha covered how to run your first Slack project successfully, because there are (surprisingly) companies who just don’t take their credit card and register the whole team, but want more structured approach (speak with me, please). And it goes again back to adoption and what it requires, changing ways people work, whom to involve in the project, differences between editions, and what to do after launch or what metrics to track.

Jason spoke about Industry/ies Clouds and how they different to the „core“ – different (higher) pricing, functionalities aligned to the industry, additional objects, OmniScript, etc. The key decision is at the beginning, check whether Salesforce covers your industry and whether the processes align with your processes and then (maybe) go for it. Be ready that it’ll limit the pool of people whom will be able to help you with the implementation and support, but it still might be worth it.

I just spoke about this picture with a client last week – they don’t want to develop, they want to click, based on experience from their old system. But it might not be the best approach long term and Tomas nicely covered 6 other myth as well. Because all of the out of the box features are great, only if they fit you. It might boost your productivity but it is still just a tool, it won’t fix everything magically on its own. It has robust instrastructure but it doesn’t mean you can use it for everything.

Evan’s presentation has been mentioned the most at the feedback we received. And while the presentation alone is pretty dry and requires a lot of your imagination, in the notes on the very last slide you have a link to video of what he achieved. Interesting idea to step out of dashboard and create lightning page showing reports charts, some flow to control refresh and a few other things to store the configuration. Not as user friendly to configure probably, but you can fit way more reports than you can to dashboard and also more filters. It might be solution for some use-cases, especially if you don’t want to go to Tableau or anything else.

Piotr and breaking best practices, topic I covered at ArchitectDreamin as well and we had great debate around it. Do you really need to be able to configure everything? What is the impact on complexity, understandability? Are hard-coded values really bad? Is duplication of code ok? Do I need a constant for everything? It resonated so well with me!

Aníbal and the „new“ approval processes – again I’m missing the demo but looks like flows are almost on par with the old-good approval process which was pretty easy to setup. I’m doing this thing so rarely that I didn’t bother with it yet.

Michal spoke about junior architects and how we still don’t know how to build them (while we know how to restore dinosaur). Presentation full of funny pictures where was probably great voice over, but this slide covers the most crucial things I would say.

Closing notes

And then Martina and her burn out topic which resonated with the audience even better than I predicted. Because the quality of our NO gives value to our YES!

Swag, main prize, food, drinks, talks – another great year. I still have no clue what I’ve been doing during the day that I didn’t have time to watch any session, but it flew by and I enjoyed it anyway.

Hope to catch you next year!

Todo’s for you:

• check the presentations
• check the photos

Todo’s for me: check the following in more depth

• Todd’s presentation
• Nebula Logger
• sfdx-hardis
• Grafana
• My Org Butler

After reading all 1078 pages of the latest release notes I realized the truth – there is happening way to much to keep up with everything. A while ago I still had a feeling I more or less knows the whole platform or at least have a clue of the different clouds. With the latest release I realized there are tons of new products I haven’t even heard about – Agentforce IT Services, HR Services and tons of other. Also it looks like three releases a year isn’t enough anymore and a lot of products switched/introduced a monthly releases as well, with the AI topic is being updated on weekly basis. Uff, I don’t really have energy to keep up with that.

Anyway, still a lot of interesting things to digest from the release notes:

• Chatter is off by default in new orgs, at the same time Slack is automatically provisioned including Slack Channels being enabled for the key objects and lovely Slack pannel added to Lightning Pages – I have to admin it looks really great. At the same time, if it is auto enabled how to handle clients who already have slack?
• Chat transcript is being moved to AWS while a lot of other things are migrated from AWS to Salesforce – not sure about the direction here
• Chat in Agentforce available in Slovak language, while Agentforce Voice in Czech
• Prompt Template deployable via metadata
• Report and dashboard brand palette
• Batch size can be specified for scheduled flows
• Email template reference by name not id which should result in easier deployment
• Hiding flow path to gain more screen space
• Date operators in decision logic in flows!
• Data table show readable record name, not id only
• Radio button groups to vizualize radio buttons horizontally
• Formula can be typed directly in actions input, no need to create a variable for it, same for transform & apex-defined variables in flows
• Timeout path for async paths to handle errors – it never occured to me so far!
• Flow version comparisons tools
• Group approval where all needs to approve – so far the flow approval process allowed only one from group to approve
• Excel Agent for auto-fill excel files (no need to export report and manipulate) (part of Agentforce Operations product – one of the new products I haven’t heard about)
• Sharing with queue member superior can be switched off – yes!
• Manage Shared List Views & Shared list view editing so you don’t have to give users the right to manage all public groups, but only those shared with their role/group, which might slightly limit their rights
• Setup with Agentforce generally available, each org can enjoy 500 free monthly actions (contractual limit, so shouldn’t pay anything above)
• Field service scheduling console looks really nice

• Embedding of existing React/Angular/Vue/JS apps in SF (developer preview) including data exchange
• Apex database operations run in user mode by default (breaking change in API 67), WITH SECURITY_ENFORCED is removed, also WITH SHARING is set by default (while apex triggers run with WITHOUT SHARING always)
• Elastic limits for queueable and future can be enabled in Apex settings and will schedule jobs over limit in throttled rate – cool, we spoke about the challenges around these jobs at ArchitectDreamin and this might be very handy feature
• Apex Anonymous Code Execution from managed packages is blocked – I know about some, which used this handy feature
• Multiline strings in Apex utilizing triple apostrophes to start/end the text – this is absolutely awesome
• API version lower than 40 is being retired (and SOAP login call all the way to version 64)
• Data Mask available for all editions? That would be great
• 10GB size limit for files uploaded to SF + malicious files detection out-of-the-box
• Get Scheduling Recommendations with Appointment Insights alias why we are getting so little slots answers

• Guide Students to Conflict-Free Course Schedules
• Elevate In-Person Meetings with Mobile AI Transcription – hard to figure out whether this is free for everyone or not, but looks pretty cool
• Log In to the Salesforce Mobile App with Email by Default, no need to enter password
• Keep Approvals Moving in Slack for advanced approvals or flow approval processes – at first glance it looks like a feature in Revenue Cloud only, but maybe it is also for the flow approvals
• Login with username first only (password/passkey as a next step)
• Find Case Experts Instantly with Agentforce Actions
• Illustrate Complex Customer Issues with Rich Text Case Descriptions (Beta)
• See Deal Engagement at a Glance in Pipeline Inspection

Looks like another busy release, with a lot of happening in flows, first release I can recall which has breaking changes in Apex and some interesting features across the product.

What others found interesting

• Salesforce Ben
• iTechCloud
• Ascend Technologies
• sigma Innovations Unlimited
• Salesforce Dictionary
• Maintask
• Salesforce Release In a Box
• Salesforce Developers Blog

A long time the <packt> publishing offered all of their books for really good price. I bought them all but didn’t really have time to read them for a long time, but finally I started one by one.

The End-to-End implementation somehow made it on top of the list and I really enjoyed the book. Kristian has really similar approach to implementations and the book resonated very well with me.

14 chapters, each with intro and summary, covering the whole process. Which activities, whom to include, what to take care about. I feel it is almost aimed at the companies implementing internally, to people without Salesforce experience. Some of the steps/tasks/chapters are so obvious, at the same time I feel they can bring a real value to someone at the beginning of their journey.

• overview of current business capabilities – I always hated this diagram so much yet understand it can bring an immense value. The tricky part for me is, that every single one of us can have a different view on the situation, can group and name the capabilities differently and it was always hard for me to come with something I would be happy with, something which won’t be obvious to everyone since beginning but would bring some new insights – but that’s probably not needed at all;
• list of delivery methodologies – I still struggle with the subtle differences between Agile, Scrum and Kanban and probably like some bits from each of them. Obviously I won’t be a great project manager;
• change managements frameworks – important part of every project which is being ignored on every single project I’ve been part of and some simplified into „management will send an email“ (in its best form). Seeing three frameworks listed in the book I want to learn more and while there is this module on Trailhead I don’t feel it gave me enough;
• KPIs or how we will spot successful implementation – very often I have to wonder how the implementation itself might improve those KPIs -> I do understand lowering processing time, but generate more leads?
• I’ve never heard of Conway’s law which says that companies do mostly lift and shift into new system, unless provided external input -> strong case (according to Kristian) to involve an implementation partner. Sadly I remember the implementations where the analyst just asked how the processes work now and didn’t care to suggest any possible improvements;
• a user story should be linked to business objectives, should have a purpose bigger than moving the process into its next step, they shouldn’t describe the solution but rather the need, problem and outcome (YES!). And another book I should prioritize – Salesforce Business Analyst Handbook;

Oftentimes, experienced functional consultants are asked to act as business analysts to run workshops, illicit requirements, and challenge the organization’s assumptions and old way of working. However, solid business analyst skills are distinct from declarative development on the Salesforce platform. While some functional consultants are great business analysts, not all are.

• deployment plan should include not just the pre/post deployment steps (still can of weird we need them these days when automated pipelines should handle everything) BUT also stakeholder communication plan – back to change management and the whole „project awareness“ topic which is hard to see in the wild. And I spoke with Ceska Sporitelna (in Czech) it is super hard as well as people don’t have really time to grasp any extra communication and are very often surprised later on that some features already exist and they don’t have a clue;
• speaking about communication, learning and documentation that’s one of the hardest topic I ever faced – as I always acted on the partner side, this was the typical area the customer didn’t want to pay for and rather do it themselves (but never had capacity to do so). Seeing what „training“ might/should include as everyone has different affinity to different learning methods gives me so many ideas, which would make the training the most expensive part of the delivery – workshops, videos, cheat sheets, one-pagers, FAQ, in-app guidance and all of that grouped by different personas and constantly updated. I probably know why at IrishDreamin the ClickLearn application won the demojam;
• when to transfer from hypercare to ongoing production support? As a partner we push into as soon as possible, seeing these four questions (have you aligned your processes, is the team trained and onboarded, are you reaching adoption targets, are you begin to achieve the business results you laid out) it should take way longer than what we typically aim for. But it is back to the KPIs you want to achieve, some of them are usually pretty long term I would say. And Kristian says kind of the same in the summary of the chapter, speaking about 1 – 12 months to begin achieving the results, not sure we should wait with go-live for that long;

Shield the development team from the Steering Committee meeting

• the continuous improvement phase after go-live and consideration to modify the partner model – different partner (due to different knowledge/experience), in-house combined with partner (and where is the line), full in-house or anything completely different;
• how to translate new feature requests into existing capabilities of your implementation and how to handle (limit/promote) the fact that as people start using the platform and learn more about the possibilities (out of the box, configuration, development) they will utilize the platform more;

All in all such a great book to learn how to deliver product into a company (yours or someone else). While there are references to Salesforce I don’t feel it is related only to SF, can be rather used by anyone.

Pokud jsem měl na začátku připravenou osnovu a základní otázky, tak to celé šlo z okna hned po úvodním představení – o AI jsem se chtěl možná tak jenom lehce otřít a nakonec jsme nemluvili skoro o ničem jiném. Což jasně ukazuje jak moc se ten svět posunul a jak moc mi asi ujel vlak, když si od AI nechám občas jenom něco poradit, ale neúkoluji ji jako svého kolegu.

Zápisky:

• Role vývojáře se mění na „orchestrátora agentů“
• Znalost platformy zůstává klíčová, i když píše agent
• Výstupy agentů vždy vypadají dobře – i když jsou špatně
• Paralelní práce s agenty je vyčerpávající
• Agentovi nelze věřit jako programu
• Jak chceš vyvíjet, jaké knihovny použiješ, jak bude vypadat dodávka – tohle si musíš vyřešit na začátku projektu. Potom to ovlivňuje úplně všechno.
• Správa kontextu a tokenů bude klíčová disciplína
• Juniorní vývojář není mrtvý – záleží na mindsetu
• Headless Salesforce je boj o přežití, ne nová příležitost
• Agentforce není mrtvý koncept – ale není pro každého. A někdy je to přesně to co potřebuješ
• Custom MCP server/end pointy jako cesta ke „snadnému zabezpečení“, protože securita bude klíčová a dost často je ohromným technologickým dluhem

But the real question is what I took away, besides the great chats, being able to speak with HR person who sees the market from outside, meeting bunch (like a lot) of new people I don’t remember seeing before and hearing a great feedback about how welcoming the community is and how even the new joiners weren’t afraid to jump into the debate and ask questions. Wow!

You can find the recordings (and second part) on Youtube and I shared the presentations on Google Drive.

Dina McLaughlin – The Platform Has Moved on – Have You? Pivoting in the Age of Agentic AI

What a great opener for the whole evening. Dina nicely summarized what happened on the market/in the industry, how it changed her life, how it is all quicker. At the same time we are moving from detesministic to probabilistic systems, where we cannot really trust or predict the output. Our roles will be more about orchestrations, reviewing and guiding, the human-in-the-loop critical due to context, creativity and bias awareness.

AI is already changing day-to-day work – for example junior devs asking AI instead of seniors, which is great and bad at the same time (cannot find the article where I found how crucial it will be to keep the human interaction and experience in the loop as well).

• The pace of change is no longer linear — it’s exponential and hard to keep up with.
• Consultants face the hardest shift → need to become transformation advisors, not just implementers.
• Businesses are confused about AI (what to automate vs what needs AI).
• Rise of autonomous agents + voice interfaces → interacting with systems without UI.
• Strong emphasis on continuous upskilling – learning is now lifelong and non-negotiable.
• Discussion around middle management potentially shrinking (AI replacing coordination roles).

Michal Verner – The Developer Who Wasn’t There: Autonomous Salesforce Development with Claude Code

Can an agent take a user story and deliver a full feature (code → tests → UI → PR) autonomously?

• AI can orchestrate the full pipeline, but only with heavy upfront setup (skills, rules, MD file, tooling) – and I had to laugh here, because it can generate it all but at the same time it expects it from you and you can get it from Reddit or other places, so now we will spend time googling skills and trying the best one just so the AI can generate the things better and quicker at the end (or maybe not)
• I need to learn new words, to fully understand what it is doing – Unravelling, really?

• The cloud.md file is the real “brain”: acts as instruction manual + standards, must be continuously refined, you “earn it” over time through iteration
• quality is inconsistent – sometimes over-engineered (too many guards, extra code), requires review or additional agents for validation
• Interesting shift in effort:
  Today: time spent coding
  Future: time spent designing instructions, pipelines, guardrails
• The agent doesn’t “understand everything” – it searches relevant files on demand, not full context; surprisingly effective even on large codebases
• Dev behavior shift: “I’m not really writing much code these days” → developers moving toward guiding, nudging, reviewing
• it was interesting to me that those agents took the whole presentation (almost 30 minutes) to generate and deploy the code, I always have a feeling the „animation“ of how it write the text is just to slow us down and it can generate the whole code base in a blink of eye

Overall feeling in the room

• Curiosity → “does this actually work?”
• Surprise → “okay, it kinda does…”
• Skepticism → “but is this maintainable / safe?”
• Humor → jokes about “agent replacing husband” / rubber duck debugging
• Realization → this is powerful, but not plug-and-play

Aleš Remta – From 75% Coverage to Sustainable Apex

• Sustainable Apex = long‑term velocity, not short‑term hacks
• Good code must be understandable, changeable, reliable, and modular – clear boundaries and contracts reduce cognitive load and prevent ripple‑effect bugs when modifying code
• Tests are not for coverage – they’re for proving requirements
• Tests are the best design tool – Writing tests first forces better API design, clearer interfaces, and more modular architecture
• Writing tests last leads to brittle, implementation‑cementing tests – When tests are written after the code, they often mirror the implementation too closely, making refactoring painful and discouraging change
• Test‑first exposes requirement gaps early – Writing tests before implementation reveals missing or unclear requirements before code is written, reducing rework and surprises at the end of development
• Use wrapper objects (e.g., DiscountRequest) to stabilize interfaces – Passing primitives or SObjects directly leads to exploding method signatures. A wrapper object keeps the interface stable even as requirements evolve .
• Guiding principles: be intentional, strict with API contracts, write more tests, and write them earlier

Dominik Hlaváč – MCP Servers: Elevating Salesforce Development

• Why MCP matters: AI needs context, and MCP provides it – compares MCP to a “USB cable for AI” that lets the model see your files, metadata, and org context – something base LLMs cannot do
• MCP architecture: client MCP server tools/resources – The server exposes tools (APIs, local files, custom logic) that the AI can call. The AI “remembers” available tools and uses them to act.
• Salesforce DX MCP server wraps the CLI into tool groups – It exposes org management, metadata operations, test execution, LWC tooling, and DevOps capabilities
• Ecosystem is exploding: Figma, GitHub, Atlassian, many more – MCP servers now exist for design, code, documentation, and workflow tools. The official registry is still in preview.
• Real use cases: dependency mapping, scratch‑org prototyping, impact analysis – MCP can automate org introspection and selective deployments: “List the org and see the whole trajectory of some attribute…”
• Limitations today: models can be lazy, inefficient, or loop endlessly – who knows how many API tools your „simple“ query will take at the end?
• Security & pricing concerns: API safer than chat, local models possible, MCP likely to become paid – “Running the API is much safer.”

And the group discusses future pricing, API consumption, and guardrails. 9pm and people are slowly heading out of the building, 10pm and we finally closing the day. Thank you all!

Podcast jsme plánovali dlouho, ale nakonec vznikl – jako většina skvělých věcí – v podstatě ze dne na den. Ohromné díky Spořce za jejich interní studio, které se postaralo o skvělou kvalitu zvuku a my jedno brzké ráno věnovali povídání s Martinem Vágnerem a Michalem Řehákem. Čas byl omezený a tak mám pocit, že je to asi jeden z nejrychleji mluvených podcastů co jsem kdy natočil, tak případně si to zpomalte.

Zápisky

• Salesforce ve Spořitelně není „jen CRM“, ale celý ekosystém, který propojuje sales, service, marketing, data i interní procesy.
• CRM v bance neslouží jen k evidenci klienta, ale hlavně k řízení interakcí, přípravě na schůzky a podpoře poradenství.
• Velké organizace netrpí technologií, ale hlavně procesy, prioritami a stakeholder managementem.
• Procesní změna je často náročnější než samotná technická implementace.
• Cílem není přepis starého systému, ale zjednodušení a inovace procesů, ideálně jejich úplné odstranění.
• Vendor lock-in existuje vždy, ale důležité je vědomě ho řídit, maximalizovat jeho hodnotu, držet systém standardní a udržitelný. Největší riziko není vendor lock platformy, ale závislost na jednom dodavateli lidí/know-how.
• AI už dnes pomáhá (např. knowledge chatbot), ale největší bariéra je regulace, ne technologie, plně autonomní poradenství je zatím legislativně problematické
• Budoucnost: AI asistenti pro bankéře, kteří připraví schůzku, doporučí produkty i styl komunikace.
• AI není jen o byznysu – má velký dopad i na vývoj – urychluje prototypování, mění roli seniorních vývojářů, vyžaduje nový mindset. Ne každá AI dává smysl – je potřeba oddělit „fancy“ use-cases od těch, které skutečně přinášejí hodnotu.
• Největší problém velkých transformací – adopce uživatelů a komunikace změn, ne samotná technologie. I při skvělé adopci platí, že uživatelé často neví, že funkce existuje, pokud není dobře komunikovaná. A i potom.
• Klíčový faktor úspěchu není jen technologie, ale lidi, kultura týmu a schopnost spolupracovat napříč organizací.

And while I do use AIs and LLMs for some tasks (like podcast summaries) I’m not really a super heavy user who would understand what they really want to do with that. Meaning it didn’t really have a high priority on my tasks list, but eventually I got to it.

I hoped that Sally would show me everything in her video, but it was just first part, developer heavy and I didn’t get much out of it.

Luckily there is a documentation – which I found before – and I went through that, did all the settings and failed. Hence this article, where I will share the updates I found and maybe explain why I feel it makes sense for regular users to connect it as well.

Step 1 – Enable MCP server

When you google how to connect Claude with Salesforce you won’t find this article, but without it the authentication will fail with super weird error 405 method not allowed.

Let’s start with that and enable MCP servers in Salesforce.

If that sounds like some mambo jumbo to you I’m not surprised, but somehow behind the scene when AI asks Salesforce when it can find something or what fields are available it will give it the answers. As you can see on the screen and in the relevant article, there are plenty of different servers/end points and you enable the ones you like. I especially like the sobject-reads which will give you read only access, no worries AI will mess anything in your org.

Open the record you like and hit the Activate button and step 1 is completed.

Step 2 – External Client App

The relevant article is easy to follow, but when you set it as specified it won’t work – some scope is missing. After some googling around I found out which one – the mcp_api is crucial one!

Save, wait some time to propagate everything across servers and get the Consumer Key, you will need to add into Claude.

Step 3 – Connect Claude

Surprisingly or even maybe shockingly, Claude doesn’t have a dedicated Salesforce connector, so you need to add it manually. Good news for bigger companies – the Claude Owner/Admin can add it on the org level and users will just connect it with their own credentials.

If you want to do it manually just follow the respective article, this one works. Actually it might be better to copy the URL from the Step 1 as it is listed there clearly (and have different URL at the same time).

Settings, Connectors, Custom Connector, ugly looking URL at the of which you will add the server you like and enabled in Step 1 (no need if you copy the URL from the Step 1). In the OAuth Client ID you will paste the Consumer Key from previous step.

Success, hopefully. If not pay attention to the URL of the error page, most likely at the end you can spot what is wrong:

• app doesn’t exist means wait a bit longer/double check the Consumer Key
• invalid Scope – most likely you didn’t add all the scopes, including mcp_api
• 405 method not allowed – most likely the MCP server is not enabled.

Step 4 – Use it

I suppose you might struggle how to use it and I was the same. But then I came with simple use cases, it gave me some data which looks relevant and save some time.

For CzechDreamin I wanted to analyze who are our top speakers – meaning they spoke the most. We have every single one as a lead in our Salesforce. As a lead for every single year they submitted a topic/spoke. When Data Cloud 360 came we tried to unify them and while it kind of worked it needed some admin work for setup and wasn’t really user friendly to get this statistic. Running a report might work but it is a lot of clicks -> Claude to save me!

Can you analyze all our leads of type Speaker and tell me which email address is the most common one (top 10). Make them unique by year (count only once per year). Count only those accepted (or in working status), basically anything besides declined. Can you also parse https://czechdreamin.com/2024/ and https://czechdreamin.com/2025 and extend it with speakers from these years (map them across the years also by name in its fuzzy form)

I can imagine you will not want to get my requirements in such form, but Claude managed. A lot of thinking from its side, a few clicks to approve their approach and what they will query from Salesforce and we are done. I must admit this conversational analysis is super cool thing!

Key findings after adding 2025:

• Aaron Crear remains #1 — now confirmed across 5 years (2019, 2020, 2023, 2024, 2025)
• Melissa Shepard, Katka Vokrinkova, Mark Jones, George Avilov, Pei Mun Lim, Todd Halfpenny, Hanna Nimchuk, Lukas Vavrin, David Fernandez Rivero all gained a 2025 year, jumping to 3–4 total appearances
• Several 2025 speakers (like Andrew Fawcett, Markus Fröhler) appear website-only with no SF Lead record — worth checking if they should be created

I must admit that I start seeing the value and maybe being braver and allow the write access as well so it can create tasks for me, update opportunities etc would be worth it.

Check my salesforce and tell me how many opportunities we have for this year

You have 3 opportunities for 2026, all currently in the „Invoice sent“ stage with a 95% probability. Here’s the breakdown:

Total pipeline value: $11,200 — and since all three are at 95% probability, you’re in great shape to close them. Would you like to dig into any of these further?

can you mark all the opportunities you listed above as Closed Won?

I have the 3 opportunity IDs from earlier. Let me update all three simultaneously.All 3 opportunities have been marked as Closed Won Total closed: $11,200. Nice work!

actually create a follow-up task for me in a week time to check whether they have all information for each of those opps.

Good, got your user ID. Now creating 3 tasks — one per opportunity — due in a week (April 28).All 3 tasks created and linked to their respective opportunities, due April 28

They’re assigned to you (Martin Humpolec) and will show up in your SF task list and activity timeline on each opp.

I always have a feeling it takes for ever, but it is probably quicker than open Salesforce and go one by one or even mass update from a list view. Yes, they all have been updated to Closed Won a few seconds later and tasks have been created.

And I’m probably sold, this conversational work – which compared to my previous experience is really spot on and doesn’t need to constantly confirm that it found the right records and everything – makes it easy to work, maybe instead of typing use voice and much more.

I’m pretty sure I’m just scratching the surface of what is possible and you do on daily basis – will you share some example how you utilize it or what you feel it will allow you to do?

A few of my notes:

• “Accidental admin” is the norm
  Many people don’t plan a career in Salesforce — they grow into it by solving real business problems, which leads to a very practical mindset.
• Hands-on experience builds credibility
  Deep, practical knowledge helps you:
  • understand system behavior
  • speak confidently with tech teams
  • avoid being purely theoretical
• End-to-end thinking > platform thinking
  A strong BA looks beyond Salesforce:
  • understands integrations (e.g. SAP)
  • sees the full business process
  • avoids siloed responsibility
• Users don’t care about systems — only outcomes
  They interact with one interface, but expect everything behind it to work seamlessly.
• Delivery often fails before it starts
  The real issues are:
  • poor data
  • unclear processes
  • missing ownership
• “You can’t automate what you can’t trust”
  Bad data = unreliable automation → validate data readiness early.
• AI / buzzwords must be translated into outcomes
  “We want AI” → define:
  • measurable goals
  • specific capabilities
• Start with enablers, not shiny features
  Before advanced solutions:
  • fix data
  • clarify processes
  • ensure visibility
• Success metrics are often missing
  Teams build and move on without checking impact → no learning loop.
• Curiosity builds trust, not resistance
  Asking thoughtful questions (instead of pushing back directly) leads to better collaboration and better solutions.
• Stakeholders must co-own solutions
  Involving them early → stronger adoption, better outcomes, and internal champions.

Check the recording on your own! And presentation if you wish to just quickly skim through.