I was in a demo last month.

Revenue management. Sales. Ops. The people who own pricing for billing, the people who live in spreadsheets, the people whose entire job is querying data and turning it into reports. A room full of subject matter experts who have spent years building institutional knowledge that only exists in their heads.

I showed them three things: AI updating a standard operating procedure in seconds. AI writing a user story from scratch. AI querying a client pricing database in plain English — no SQL, no ticket to the data team, no waiting.

The room went silent.

Then someone said: we need to put controls around that. We can't just have people querying.

I had to talk them off the ledge. Entitlements. Parameters. Not everyone gets access — we can build a wall. It took a minute.

Afterward, my boss pulled me aside. I don't think you registered their faces. They weren't pushing back because it was bad. They were scared because it was real. They saw something in that room that I had already absorbed so gradually I'd stopped noticing it.

They saw their promotion to customer in front of them.

What AI Access Is Actually Worth

Here's the thing I've been sitting with since that demo: the people in that room aren't wrong to be unsettled. But the threat isn't what they think it is.

The threat isn't that AI replaces them tomorrow. The threat is that someone who knows how to build with AI — really build, not prompt-and-pray — walks into a non-technical space and starts automating the things those experts have been doing manually for years. And that person doesn't need to be a senior engineer. They just need to know how to build workflows. How to wire agents together. How to take a business problem and close the gap between idea and output.

I call it the builder role. And it's already forming.

Right now it's mostly product and tech people. But the next wave isn't a software engineer replacing a data analyst. It's someone who understands the business and knows how to build — moving into spaces that have never needed that combination before. HR. Operations. Revenue management. Legal. Anywhere there are manual processes, institutional knowledge locked in people's heads, and no one on the team who knows how to automate it.

The Meta layoffs this week — 10,000 people. Those aren't people disappearing from the workforce. Those are builders, freed up, looking for somewhere to apply what they know. They're coming for roles you wouldn't expect. And they're going to be very, very good at it.

The Pipeline I Didn't Build Alone

My tech lead — I'll call him Matt — and I have been building something together. What started late last year as a way to write better user stories has evolved into something much bigger: a full idea-to-code pipeline.

Nine agents. One workflow.

A half-formed thought drops in — messy, unstructured, incomplete. The intake agent reads our entire knowledge brain first: business rules, prior stories, team conventions, feedback history. It asks only what it can't already find. The spec agent generates a PRD, a technical spec, and a test plan from a single input. The architect agent scores confidence against the live codebase and routes accordingly — high confidence goes straight to code generation, low confidence escalates to Matt with specific, answerable questions. The coder agent generates Java implementations and unit tests. The testing agent verifies every acceptance criteria before a PR ever opens. The PR agent packages it, tags it AI-generated, and submits it with the full audit trail attached.

The brain updates after every merged PR. It compounds. It gets better the more we run it.

I told Matt this week: in a year, when this is ironed out, I don't see why the two of us can't build everything on our own.

He didn't disagree.

What Made That Possible

GitHub Copilot. Unlimited tokens. No hard limits. No approved-use-only restrictions. No watered-down model with the useful capabilities stripped out.

I want to be specific about this because I think most people undercount it when they're evaluating a role.

Outside of work, AI access costs real money. Claude has rate limits. Every additional capability comes out of your pocket. If you want to build seriously — not dabble, but actually build agents and pipelines and systems — you hit the ceiling fast. The metered version of this is genuinely limiting.

At work, I have GitHub Copilot as my bread and butter — running on Anthropic's models, no cap on what I can do with it. An internal LLM. M365 Copilot. AI natively embedded in half the tools I touch. It's what I use to build the pipeline, write the stories, run the demos, do the work.

And here's the part that matters: every hour I spend building at work is an hour I'm sharpening skills I own. The company keeps what we built — the system, the pipeline, the code. But the way I think about problems now, the ability to look at any business process and see the agent architecture underneath it, the instincts I've built for how to close the gap between idea and output — that lives in my head. It walks out the door with me.

I am learning to build on my employer's dime. At scale. With no token anxiety.

What I Ask In Every Interview Now

When I'm evaluating a role, AI access isn't a nice-to-have. It's a requirement. I ask about it the same way I ask about scope and comp.

What models do you have access to? Are there token limits? What are you allowed to use it on? Do you have a watered-down internal version with half the features locked? Is Copilot available, or are we on something proprietary that nobody knows how to use?

The answers tell me everything.

A company with hard token caps and a restricted approved-use list is telling me something. They're telling me they're managing AI like a cost center instead of treating it like infrastructure. They're telling me the builders on their team are working with one hand tied. They're telling me they haven't figured out yet what this is.

A company with generous AI access and a culture that says go build something is a different conversation entirely.

I'd take less base salary for the right AI access. I'm being dramatic — slightly. The real point is that generous AI access quietly inflates what an offer is actually worth. Most people aren't doing that math.

Tokens are currency right now. Expensive currency. If your employer is handing them out without a meter running, that's compensation. Treat it like it.

The Room That Got Quiet

I keep coming back to that demo. The silence. The instinct to shut it down before they'd even processed what they saw.

Those weren't bad actors trying to block progress. Those were smart people whose expertise had always been the asset — and who recognized in real time that the asset was being repriced.

The SOP that used to take a senior ops person two hours to update: seconds. The database query that used to require a ticket and a three-day wait: plain English, immediate. The user story that used to require a half-hour refinement meeting: drafted, structured, ready for review before the meeting starts.

The knowledge doesn't go away. But the bottleneck does. And for people whose value was the bottleneck — being the one who knew how to do the thing — that's a real shift.

Here's what I want to say to the people who got quiet in that room: the answer isn't to put a wall around it. The answer is to be the one who knows how to use it. To learn the business and learn to build. To become the person who walks into a space like yours and automates the right things — with you, not instead of you.

The builders are coming regardless. The question is whether you become one.

What This Actually Looks Like in a Comp Negotiation

For the HENRYs reading this who are in the market or thinking about it: AI access belongs in your total comp conversation. Here's the practical version.

Ask specifically. "What AI tools do your teams have access to, and are there token limits or use restrictions?" A vague answer — "we have some AI tools" — is a red flag. A specific answer that includes model names, "unlimited" access, and encouragement to build is a green flag.

Price the gap. If you're leaving a role with unlimited Copilot access and the new role has none, you're taking on a real cost — both in dollars (if you rebuild that access personally) and in compounding (what you won't be building on company time). Factor that.

A company's AI access policy tells you how seriously they're thinking about the next three years. If they're restricting it heavily, they're either behind or scared. Neither is where you want to be building your career right now.

The people who are going to be most valuable in the next few years aren't the ones with the most credentials. They're the ones who've had the most reps. Who've built the most pipelines, run the most agents, learned from the most failures.

Generous tokens are reps. Don't leave them on the table.

— The Daring Dime

SEO Meta Title: I'd take less salary for unlimited AI tokens. Here's the math.

SEO Meta Description: AI access is compensation. Tokens are expensive outside of work, and what you build on your employer's dime compounds. Here's how I evaluate it in every role.

URL Slug: ai-access-is-compensation

Tags: Salary Optimization, AI as Leverage, Career Strategy, Total Compensation, Builder Role

Excerpt: The room went silent. I'd just shown a group of business experts what AI could do to their workflows in real time. The silence wasn't confusion — it was recognition. This is a post about what that moment taught me, and why I now negotiate AI access the same way I negotiate salary.

Hero Image Alt Text: Overhead view of a laptop slightly cracked open, dark screen, next to a coffee mug and open notebook on a desk ledge overlooking a city skyline

Hero Image Recommendation: Overhead angle, laptop cracked just enough to signal someone is working — screen dark, not glowing. Dark coffee mug and open notebook beside it. Floor-to-ceiling window with dense city skyline behind. Cool, overcast light. No people, no warm amber, no glowing screens. The image that says: someone who builds is always running something in the background.

ChatGPT Image Generation Prompt: A cinematic overhead shot of a minimalist dark desk ledge next to a floor-to-ceiling window overlooking a dense city skyline, overcast daylight. A laptop slightly cracked open with a dark screen, a dark coffee mug, and an open blank notebook arranged on the desk. Cool, neutral tones. Clean, intentional composition. Photorealistic, no warm amber light, no people, no visible screen content.