Insights | The realfast view
After AI, There Is No Average
With AI, the bell curve an enterprise is built on becomes a power law: a few people produce most of the value, and most of the rest produce slop.
Insights | The realfast view
After AI, There Is No Average
With AI, the bell curve an enterprise is built on becomes a power law: a few people produce most of the value, and most of the rest produce slop.
realfast labs | From the Labs
I stopped having a job description.
There was no formal AI training, but there was the freedom to try everything and brilliant colleagues to learn from. Two years later, a team of agents I built runs People ops, payroll, hiring, marketing and writing.
Engineering | From engineering
How to estimate AI-native software projects
Estimating under uncertainty: the pitfalls, a five-step method, and what changes with AI
realfast labs | From the Labs
Every AI search tool finds something. They differ on whether it's true.
The most dangerous answer an AI search tool can give isn't "I don't know" - it's a fluent description of a company that barely exists. We built a benchmark around exactly that case, and it re-ranked every tool we tested.
realfast labs | From the Labs
Onboarding was never a knowledge problem
Two days into a new job I had access to everything and understanding of nothing - buried not by too much information but by not knowing where to look. So I built /brief: a command that queries every tool we run on and hands back one connected picture of any client, repo, or person.
Insights | The realfast view
AI doesn't have bad taste. It's just missing constraints.
Teams are shipping internal tools faster than ever, and the interfaces all look subtly off. The reflex is to blame the model's taste. The real problem is constraints, and where they now need to live.
Insights | The realfast view
The hidden cost of a round-trip
When AI compresses days of execution into hours, the day you lose to a vague email stops being invisible. Seven habits to close an issue in one message instead of three.
Insights | The realfast view
Your old core is slowing you down
Old core standing in the way of new customers? It doesn't have to be that way. Once upon a time rebuilding a legacy core meant locking your best people down on a hard job with uncertain outcomes. Not anymore. Answer five questions, put AI on the grind, and that launch date becomes a solid commitment instead of forlorn hope. This is how we do it.
Insights | The realfast view
Velocity isn't lines per day. It's knowing which lines matter.
The entire production diff was about 15,000 lines, roughly thirty lines per engineer per day. AI's real contribution wasn't writing them. It was comprehension, not codegen.
Insights | The realfast view
Your AI roadmap probably has too much AI
Most production systems should use AI at the edges and deterministic code in the core. As per-token pricing turns honest, a lot of what's on AI roadmaps today won't survive the look.
Insights | The realfast view
AI transformation patterns: the playground
If you take AI transformation advice at face value, the work looks done once the tools are bought, and the business still moves like it did a year ago. The Playground is the pattern that turns the tools into a business that moves differently.
Insights | The realfast view
Don't go AI native
Going AI Native means one thing: your customers get superhuman speed and superhuman quality no competitor can humanly match. A Head of AI and a Center of Excellence won't get you there: Parnas, Conway, and Jevons explain why.
Insights | The realfast view
There is no product
The "SaaS is dead" debate is fixated on pricing. The real shift is deeper: AI is converting one class of software after another from a scarce asset into cheap, replicable inventory, and below that line, there is no product.
Insights | The realfast view
Nobody sets out to build a context graph
We tried to fix our timesheets and accidentally built a context graph. Here's what we learned about why operational discipline matters more than architecture.
Engineering | From engineering
Three context layers every Agentforce implementation needs (and how to build them)
Salesforce spent $8B on Informatica because AI agents need unified context. Here's how to build the data architecture that makes Agentforce actually work in production.
Insights | The realfast view
What happens when you deploy AI agents without restructuring
35% of companies are already using agentic AI, with another 44% planning deployment soon. But few have restructured how they deliver projects. That gap is where implementation timelines extend and where fixed-price engagements turn into scope management problems.
realfast labs | From the Labs
What we learned deploying AI agents for TASC Outsourcing
We built two Agentforce agents for TASC Outsourcing that handled 5,100+ sales interactions and improved their email response rates from 0.33% to 1.93%.
realfast labs | From the Labs
The $21,600 Problem: Why Salesforce Developer Onboarding Costs More Than You Think
Developer onboarding costs $21,600 annually per 20-person team. See how ExoPair compresses 3-5 days of onboarding into day-one production deployments.
realfast labs | From the Labs
Nobody wants to update tickets. So we automated it
Developers hate updating tickets. Managers hate chasing updates. We use Exo, our internal AI agent, to make the whole problem disappear.
Insights | The realfast view
Meet ExoHelp: the AI that eliminates Salesforce delivery friction
ExoHelp turns 3-5 day Salesforce requirements into 30-minute sprints. See how AI analysis eliminates clarification cycles and delivers developer-ready tickets instantly.
Stories | Field notes
Why the best engineers are becoming Consultants again (They just don't call It that)
The best engineers used to be consultants who built foundational software. Then consulting became uncool. Now Forward Deployed Engineers are making it desirable again by solving real business problems instead of hiding behind JIRA tickets.
realfast labs | From the Labs
Inside look: how AI-native teams deliver Salesforce projects 3X faster
See how Salesforce work lands in days, not months.
Stories | Field notes
Building realfast: solving the last mile problem in enterprise AI and what we learned
Two leaders who scaled Gojek to 7M daily orders share why most enterprise AI fails and what actually works beyond demos. Lessons from building realfast.
Engineering | From engineering
Salesforce Agentforce flex credits: from $2 a conversation to 10¢ an action
Salesforce Agentforce flex credits changes everything. From $2 per conversation to 10¢ per action - pay only for real work done. Game changer.
Engineering | From engineering
The 25% of Agentforce Interactions That Consume 75% of Your Budget
The AI Effort Paradox: Why trying harder costs more
Engineering | From engineering
The 3X performance gap: Why your Agentforce prompts need batch testing
Discover why single prompt testing fails in production. Learn how batch testing revealed a 3x performance gap between AI models and the simple technique that boosted Agentforce speed without sacrificing quality.
Stories | Field notes
When Reasoning Becomes Productized: AI Transformation x IT Services
AI is transforming IT services' economic model by decoupling revenue from headcount
realfast labs | From the Labs
Scaling hiring with AI: How realfast uses AI Interviews to save time
We were drowning in intro calls. So, we brought in an AI interviewer to do the talking for us, literally. Here’s how it’s helping us scale faster, stay human, and still make great hires without burning out.
Engineering | From engineering
Peeking inside the Mind of the Machine
On setting up reporting for Agentforce and why reporting is the heart of AI reliability
Engineering | From engineering
Some useful best practices for Agentforce implementations
Best practices and tricks of the trade that ensure your Agentforce agents operate accurately and consistently, even under complex conditions.
Insights | The realfast view
So you think running API calls in a loop makes an agent?
New research reveals that simple AI agents fail at enterprise tasks over a third of the time. And the solution isn't just better models.
Engineering | From engineering
Going x to 10x on high-volume recruiting with Agentforce
Let's look at how Agentforce can do the heavy lifting so time-starved recruiters can focus on building relationships and making successful placements.
realfast labs | From the Labs
4 surprising things we learned testing agents on Agentforce
We ran into some unexpected quirks while testing Salesforce agents. Here's what you need to know before starting.
Insights | The realfast view
Making sense of data without losing your mind
Dashboards are the wrong interface for asking a question.
realfast labs | From the Labs
What do you pay for while using Agentforce?
While the base pricing for Agentforce seems straightforward, the real cost structure involves understanding conversations, Einstein requests, and data credits – each with its own nuances and optimization opportunities.
realfast labs | From the Labs
Why AI-first Salesforce implementation changes everything
Salesforce is a powerful tool, but traditional implementations often lead to data chaos within months. An AI-first Salesforce implementation changes that.
realfast labs | From the Labs
Async work is where the juice is.
Agents that can do work asynchronously will be crucial to make GenAI scalable and viable for enterprise workloads.
Stories | Field notes
Speed up Salesforce Implementations by 2x with Vayu
How Vayu, our agentic platform for Salesforce teams, cut the daily operational load, change-request turnaround, and testing cycles in 2024.
Insights | The realfast view
Here's how to evaluate AI Agents for Enterprise use-cases
A problem evaluation framework we've developed over the past year for both GenAI founders and investors.
realfast labs | From the Labs
Speed is the name of the game.
Speed sells AI tooling, not cost savings.
Insights | The realfast view
Dumb agents beat smart ones.
You want your agents to be just smart enough to do the task consistently and reliably, and not any smarter than that.
realfast labs | From the Labs
For AI Transformation, it helps to start small.
It actually helps to be less ambitious when starting out with AI transformation and take small, meaningful steps toward increasing velocity and throughput.
Insights | The realfast view
For Applied AI, Data = Work
Foundational AI and Applied AI operate on entirely different planes. So, the approach you take to build an applied AI company will also be very different.
realfast labs | From the Labs
Our unit test agent is now in production.
We're talking 12-16x faster test creation, with 95%+ code coverage.
realfast labs | From the Labs
Agents are minimum viable intelligences
For typical enterprise workloads, LLMs are raw material. You will need software engineering to turn them into a scalable resource.
Insights | The realfast view
Evals for agents are a hard problem
Evaluating an AI tool's performance in real-world workflows is inherently more complex than benchmarking the models themselves on established datasets.
Insights | The realfast view
Beyond wrapper tech
Wrapping an API in an interface is not the hard part.
Stories | Field notes
Hello World!
IT services is being rebuilt around agents.