This is the web version of the newsletter which you can subscribe »
Hello
This week I want to start with something a bit different - a shoutout to a project worth your attention: "Sunhive Quest". A completely free pixel-art RPG where you become the new Salesforce administrator at a (slightly chaotic) solar company. You explore the office, talk to the team, and solve their real problems - for real, inside your own free Salesforce Developer org, with the game checking your work automatically.
Every character, including your AI Coach, runs on a real AI model, so you can ask anything, in any language, and get a specific answer - like a mentor on call. It prepares you for two things at once: day-to-day admin work and the Salesforce Administrator exam - by doing, not by cramming.
55 missions across 8 acts, English and Polish. Free, no installation, runs in the browser.
As for the rest of this week's links: Salesforce officially cancelled the Permissions in Profiles retirement, which is a relief for many teams mid-migration. There's also a solid piece on why data modeling matters more than prompt engineering when it comes to AI. And for architects - a survey showing communication skills now rank above coding. Make of that what you will.
Enjoy the read.
For everyone
-
14 Ways to Make Salesforce Reports Run Faster
https://www.salesforceben.com/ways-to-make-salesforce-reports-run-faster/
Slow Salesforce reports usually happen because they're asking the system to do too much work - not because something's broken. Reports often start simple but get bloated over time with extra columns, filters, and formulas that pile up. The fix is straightforward: make Salesforce work smarter by cutting unnecessary stuff and optimizing what's actually needed. This means removing unused columns, cleaning up filters, simplifying formulas, and keeping reports focused on answering specific questions instead of trying to do everything at once. Doing this basic cleanup makes reports run faster and saves everyone time. -
How Much Does the Salesforce Community Trust Salesforce Right Now?
https://www.salesforceben.com/how-much-does-the-salesforce-community-trust-salesforce-right-now/
Salesforce is running a quick community survey until July 19 that takes about 10 minutes and lets you share honest feedback anonymously - it's basically your chance to tell them exactly what you think about their recent moves. The survey cuts right to the point with tough questions like whether your view of Salesforce got better or worse in the last year, if you actually trust them to do what's best for customers, whether you feel safe using their AI tools, and if you can keep up with how fast their products are changing. These aren't soft marketing questions - they're the real deal, suggesting Salesforce might already know the community has some concerns and wants to find out how serious things really are before making changes. -
Back to Basics: Modeling the Business, Not the Software
https://forceforfun.com/2026/07/30/back-to-basics-modeling-the-business-not-the-software/
Everyone's focused on writing better prompts for AI in Salesforce, but that's putting the cart before the horse. Before AI can do anything useful - answer questions, suggest next steps, or summarize cases - it needs to understand the actual business first. The real foundation is the data model, which is basically how the business represents itself inside Salesforce. The data model isn't just a random collection of objects and fields - it's the entire story of how the business works, what exists in it, and how everything connects. Every object created, relationship defined, and automation built adds to that story. AI doesn't create this worldview; it just works with what's already built. Great prompts help, but they can't fix a broken or confusing data model underneath. -
Salesforce Will Only Survive the AI Era If It Can Compete Against Vibe Coding
https://www.salesforceben.com/salesforce-will-only-survive-the-ai-era-if-it-can-compete-against-vibe-coding/
Salesforce has faced plenty of rough patches - stock drops, questions about its products and direction, even criticism of its CEO - but it's still a massive, successful company worth billions. Wall Street isn't totally convinced, though interest remains strong. With Agentforce developing, Salesforce could be heading up again - but only if it can beat out the rising competition from vibe-coded alternatives that are gaining ground in the AI space. -
Salesforce Cancels the "Permissions in Profiles" Retirement: What It Means for Admins and Architects
https://sudipta-deb.in/2026/07/salesforce-cancels-the-permissions-in-profiles-retirement-what-it-means-for-admins-and-architects.html
Salesforce just cancelled its big plan to force everyone off managing permissions through Profiles - something admins and security teams have been stressed about for years. The company was supposed to push everyone toward Permission Sets instead around Spring '26, but customer feedback and some missing features made that impossible. So now organizations can keep using Profiles if they want, which is honestly a relief for teams that were already deep into switching strategies or worried about the disruption. -
Why Is Salesforce Data Cloud Still So Difficult to Implement?
https://www.salesforceben.com/why-is-salesforce-data-cloud-still-so-difficult-to-implement/
Salesforce's Data 360 has been super important to the company, but it keeps getting renamed so often that nobody really knows what it's supposed to do anymore - each rebrand puts it in a different spot in the business setup, which just makes things more confusing. All those changes have frustrated users while also showing how the product keeps shifting its purpose, making it one of the hardest Salesforce tools to actually figure out and use. -
Designing Salesforce Integration Users for Enterprise Security and Scale
https://sforcemaximizer.com/designing-salesforce-integration-users-for-enterprise-security-and-scale/
Salesforce teams often debate whether integrations need separate Profiles or just Permission Sets - but that's actually the wrong question to ask. The real issue is deciding what external systems should be allowed to do inside Salesforce in the first place. Instead of focusing on implementation details like Profiles and Permission Sets, architects need to build a security model that can handle dozens or hundreds of integrations while staying secure, auditable, and flexible. It's about trusting the right systems with the right access - not just handing out permissions. -
Is Hardcoding Ids a Good Idea in Salesforce Flows?
https://salesforcebreak.com/2026/07/22/hardcoding-ids-in-salesforce-flows/
Hardcoding Salesforce record IDs directly into Flows might seem like a quick fix - copy the ID from the URL, paste it into a text variable, done - but it's a risky move that causes major problems. While it works fine in sandbox, deploying this approach to production often breaks things because IDs change across different environments. This anti-pattern creates fragile automation that fails during deployments, especially when routing emails to queues, assigning tasks to system users, or referencing specific Record Types. Understanding why this happens and using proper alternatives instead keeps automation stable and reliable across all Salesforce environments. -
Apex Integration Testing in Summer '26
https://bobbuzzard.blogspot.com/2026/07/apex-integration-testing-in-summer-26.html
Salesforce Summer '26 quietly introduced Apex Integration Testing - a new feature that finally lets developers test integrations properly without messy workarounds using external tools and fake mocks. Right now it's pretty limited - only available in Developer Preview for scratch orgs and works with Data 360 and Agentforce services - but hopefully it'll expand soon to support any authorized endpoint. This could be a game-changer for how testing gets done on the platform, moving away from those clunky makeshift solutions people have been cobbling together.
Job Market
-
Are Salesforce Events Really Dying Off?
https://www.salesforceben.com/are-salesforce-events-really-dying-off/
Salesforce's big events like Dreamforce and TrailblazerDX used to pack massive crowds with excited attendees, but things feel different now - the company's big shift toward AI and Agentforce has created mixed reactions, with some saying attendance and energy have dropped, though it's unclear exactly what's really happening at these events lately. -
How Salesforce Professionals Can Navigate Career Anxiety
https://www.salesforceben.com/how-salesforce-professionals-can-navigate-career-anxiety/
Career anxiety hits everyone - even super experienced pros in tech where pressure's always high. Salesforce professionals know this struggle well, whether it's general worry about job security or that "midlife crisis" feeling many face right now. Three Salesforce MVPs shared how they tackled their own anxiety issues, offering real advice that might help anyone dealing with similar stress.
For devs
-
How to Connect Claude with Salesforce Hosted MCP Servers
https://salesforcetime.com/2026/07/15/how-to-connect-claude-with-salesforce-hosted-mcp-servers/
Salesforce just launched Hosted MCP Servers - a game-changer for connecting AI assistants like Claude to your Salesforce org without building custom integrations or running local servers. Now Claude can directly access your Salesforce data and automation through Salesforce's own infrastructure using OAuth - meaning you can query records, run reports, and trigger flows just by talking to Claude in natural language, no more navigating Lightning Experience or waiting for developers to build API layers. This post walks through how to get Claude connected to Salesforce using these hosted servers.
For architects
-
The Most Valuable Skills for Salesforce Architects in 2026 - And Coding Isn't #1
https://www.salesforceben.com/the-most-valuable-skills-for-salesforce-architects/
Salesforce Architects are in huge demand - Technical Architect jobs jumped 27% while supply only grew 4%, and Solution Architects are up 21%. As Salesforce keeps adding AI, Data 360, automation, analytics, and integration tools, architects need to bridge business strategy with technical execution, managing complexity while helping organizations scale. The latest SF Ben Architect Survey shows something surprising - communication skills rank as the most valuable for architects in 2026, beating out coding which ranked dead last. This doesn't mean technical knowledge doesn't matter, but it highlights a growing trend - the best architects aren't just tech experts. They're communicators, facilitators, and strategic thinkers who understand how to connect different parts of an organization. -
Why Salesforce Architects Are Rethinking Their Careers
https://www.salesforceben.com/why-salesforce-architects-are-rethinking-their-careers/
Salesforce Architect jobs are booming - demand shot up 27% for Technical Architects and 21% for Solution Architects. Sounds perfect, right? Wrong. A new survey shows only 31.8% of architects actually want to keep doing what they're doing now, down from 42.6% last year. Instead, more are eyeing enterprise architecture, executive roles, or careers outside Salesforce entirely. The twist? This isn't really people bailing on Salesforce - it's the job itself getting way bigger. AI, data, and connected tech stacks are changing how enterprises work, so architects are questioning whether deep Salesforce expertise alone is enough anymore, or if they need to think way beyond just the platform. -
Building Reliable Production AI with Durable Workflows
https://engineering.salesforce.com/building-reliable-production-ai-with-durable-workflows/
Building AI prototypes is simple - just send a prompt, get a response, done. But production AI is messy. When running across thousands of records with simultaneous model calls, things break - rate limits hit, workers crash, deployments roll out mid-execution. The real challenge isn't whether the AI gave the right answer anymore; it's tracking what finished, what's still running, and what can safely retry without duplicating work. That's when debugging shifts from AI behavior to distributed systems headaches. Every successful AI platform faces this transition - moving beyond just smart responses to ensuring that intelligence survives long-running execution, partial failures, and restarts without making users start over. The key lesson: production AI relies on durable workflows that keep things working through failures, not just clever prompting. -
How AI Reduced Customer Bug Triage from Nearly a Year to Less Than a Week
https://engineering.salesforce.com/how-ai-reduced-customer-bug-triage-from-nearly-a-year-to-less-than-a-week/
Salesforce built BugWiser, an AI system that speeds up how customer bugs get sorted and fixed. Priya Sethuraman's team combined machine learning with AI to cut bug review time from over 300 days of manual work down to just a week. As Salesforce shipped features faster, handling customer feedback became messy and slow - bugs weren't getting organized consistently or acted on quickly enough. BugWiser standardizes how bugs get classified, helps teams find root causes faster, and gets fixes to customers quicker. The real win? Turning customer bug reports into actionable signals that reveal bigger quality issues while the company keeps shipping new features at speed.
Reddit Discussions
-
Is Salesforce's Salesforce the most complicated Salesforce in existence?
https://old.reddit.com/r/salesforce/comments/1ukr4g1/is_salesforces_salesforce_the_most_complicated/
Someone's wondering if Salesforce's internal org ("Org 62") is the most complicated Salesforce instance that exists - like, who else would even have the expertise and need to build something that complex? They're thinking maybe US government systems could compete, but basically asking: if not Salesforce itself, who has the product knowledge and scale to pull off something that crazy? Just some fun speculation about which real-world Salesforce setup is the most insanely complicated. -
Is Anyone Here Reinventing The Salesforce MCP Wheel?
https://old.reddit.com/r/salesforce/comments/1umqx9r/is_anyone_here_reinventing_the_salesforce_mcp/
User is asking if anyone's actually using Salesforce MCP for real, complex stuff - not just basic SOQL queries. They tried the official MCP setup before and found it pretty useless, especially for Revenue Cloud, while Claude + SF CLI worked way better. Businesses wanted MCP solutions ASAP with agents handling actual workflows, so they built a custom setup with HTML components that execute methods in the right sequence and it works great. Now they're wondering if it's worth keeping their solution alive or if they should switch to Salesforce's official MCP now that it's supposedly better - or if Salesforce's version is still half-baked. Looking for others' real experiences with advanced MCP implementations. -
We're the team that gave Salesforce's Agentforce the option to be powered by Google Gemini. Ask us about why...
https://old.reddit.com/r/u_salesforce/comments/1uqdvg8/were_the_team_that_gave_salesforces_agentforce/
The Salesforce team that integrated Google Gemini into Agentforce shared their experience building AI agents for enterprise work. While everyone's pushing "deploy AI or fail," they've learned it's way more complex than just picking a model. After a year of work automating customer service, sales, and field operations tasks, they partnered with Google to offer Gemini 3.5 Flash - launched in May and available in Agentforce by June. Testing showed impressive results: 100% task completion, 100% state accuracy, 94% action rate, and tasks resolved in 2.2 turns - roughly 3x faster than competitors. But speed isn't everything - they stress that reliability and trust matter most when agents work alone in real environments. They've hit bumps and learned what doesn't work, but found the winning formula: combining Agentforce's predictable Agent Script with Gemini's fast, capable performance to create powerful yet trustworthy agents.
For AI enthusiasts
-
Complete Guide to Agentforce Pricing Options
https://www.salesforceben.com/complete-guide-to-agentforce-pricing-options/
Salesforce recently launched a new pay-per-resolution pricing model for Agentforce, shifting how customers get billed - now you pay based on actual results instead of just usage. The Help Agent tool focuses on solving customer issues, and while Salesforce kept messaging simple by saying costs tie to "outcomes, not activity," plenty of details about how this actually works remained fuzzy. Product leadership stepped in to answer key questions about what this pricing change really means for customers. -
Why Even Anthropic Can't Vibe-Code Away Salesforce
https://www.salesforceben.com/why-even-anthropic-cant-vibe-code-away-salesforce/
Anthropic's AI is supposedly letting companies build their own software instead of buying from Salesforce. But here's the thing - Anthropic just hired a Salesforce Developer and is expanding its CRM use. Sounds ironic, right? Not really. It actually shows that AI and SaaS companies like Salesforce aren't enemies - they work together. The reality is way more complicated than "AI will kill traditional software companies." -
Workshop on Building a Deterministic RAG Agent in Agentforce
https://salesforcediaries.com/2026/07/28/workshop-on-building-a-deterministic-rag-agent-in-agentforce/
Most B2B SaaS companies struggle because support requests pile up faster than teams can handle them - so they turn to chatbots, which often backfire by making up answers, refusing to connect customers with humans, or getting stuck in loops. This post lays out a better approach: building a RAG-grounded FAQ agent in Agentforce that works the same way every time by using explicit if/else logic instead of relying on AI to follow vague instructions. The agent stays simple - it only answers questions from existing knowledge and transfers to humans when needed, with zero ability to refund, provision, or change anything. That focused design cuts out tons of messy requirements while introducing new security concerns worth paying attention to. The guide uses a fictional company called Nimbus Workspace to keep everything concrete and real, with 40+ test cases you can adapt for your own setup. -
Salesforce Agentforce Help Agent
https://www.infallibletechie.com/2026/07/salesforce-agentforce-help-agent.html
Salesforce Help Agent cuts through the headache of setting up AI agents in big companies - no more wrestling with complicated frameworks or complex technical setups. It lets organizations build autonomous support for customers and employees right inside Salesforce itself, keeping everything simple and protecting existing knowledge bases. This guide helps architects and admins get the Help Agent running smoothly in their Setup without the usual tech complications.
View upcoming Salesforce events (next 7 days)
 |
Grzegorz Skaruz
|
If you no longer want to receive messages from me, you can always unsubscribe