It’s 5:47am and I’m halfway through a set of deadlifts when my phone buzzes. Not a notification I asked for. A summary I didn’t request. Sally has already reviewed the three active client engagements on my board, flagged a dependency risk on one project that I would have caught by 10am, and drafted a pre-read for a stakeholder meeting I have at 9. I haven’t opened my laptop. I haven’t had coffee. And the most important thinking of my day is already underway.
This is what it feels like when AI stops being a tool and starts being infrastructure.
The moment it clicked
I’ve been building with AI for long enough that I thought I understood what it could do. I was wrong — or at least, I was thinking too small. For a long time, I used AI the way most people do: as a smarter search engine, a faster first draft, a way to summarise things I didn’t have time to read. Useful. Incremental. Safe.
Then I built Sally, and something shifted.
There’s a specific moment when an AI agent stops feeling like software and starts feeling like a colleague. It’s not when it gives you a good answer. It’s when it does something you didn’t ask for — and it’s exactly right. Sally crossed that line about three weeks in. She started surfacing patterns across client engagements that I hadn’t connected. She began pre-loading context before meetings that I would have spent twenty minutes assembling manually. She flagged a scope risk on a Tuesday that saved a difficult conversation on a Friday.
She didn’t wait to be asked. She just operated.
“There’s a moment when an AI agent stops feeling like software and starts feeling like a colleague. It’s not when it gives you a good answer. It’s when it does something you didn’t ask for — and it’s exactly right.”
What Sally actually does
Sally lives in Telegram, which means she’s wherever I am — on my phone at the gym, on my laptop in a client’s office, on my watch between meetings. She’s not a dashboard I have to check. She’s a presence that checks in on me.
Her core job is managing the operational backbone of my consulting practice. She maintains a living pipeline of every active engagement, every prospect conversation, every piece of context that matters. When I record a voice note after a client call, she turns it into structured notes, updates the project board, identifies follow-up actions, and routes them to the right place. When I say “new project,” a GitHub repository, a project board, and a development environment exist within seconds. Not minutes. Seconds.
She creates structured task breakdowns from ambiguous briefs. She triggers automations across tools that don’t know about each other. She maintains a persistent memory of every engagement — not just what happened, but what mattered. And she does all of this invisibly, in the background, while I’m doing the work that actually requires a human brain.
What Sally is becoming
This is where it gets interesting. Because Sally today is the worst version of Sally that will ever exist.
What I’m building toward is an agent with genuinely compounding intelligence. Persistent memory that grows richer with every engagement, every conversation, every decision. She’ll know clients before meetings start — not from a briefing doc, but from months of accumulated context. She’ll surface research that lands in my inbox before I knew I needed it, because she understands not just what I’m working on but how I think about it.
The gap between idea and execution is collapsing. The overhead that used to make small consulting practices slow — the admin, the project setup, the context assembly, the follow-up tracking — is evaporating. What’s left is the thing that clients actually pay for: senior thinking, applied to their specific problem, without the drag.
“We are not optimising workflows. We are rebuilding how a consulting practice operates from the ground up.”
What this means for clients
Here’s the part that matters if you’re reading this as someone who hires consultants, not someone who builds AI.
When Sally handles the operational load, you get a consulting practice that operates like a product company. Fast. Structured. Always on. The turnaround between “we discussed this on Tuesday” and “here’s the structured analysis” isn’t days — it’s hours, sometimes minutes. The context from your last engagement doesn’t live in someone’s memory and a folder of old documents. It lives in a system that compounds and surfaces it exactly when it’s needed.
You get the depth of a senior operator with the responsiveness of a team. Not because I’ve hired a team. Because I’ve built one.
The bigger picture
I want to be direct about what I think is happening, because I don’t think most people in professional services have fully processed it yet.
We are not optimising workflows. We are not adding AI features to existing processes. We are rebuilding how a consulting practice operates from the ground up. The firms that figure this out in the next eighteen months will be structurally different from every other firm in the market. Not incrementally better. Not faster at the same thing. A different category entirely.
The economics change. The speed changes. The depth of service changes. The relationship between firm size and capability changes. A solo practitioner with the right AI infrastructure can now deliver what used to require a team of six — not by cutting corners, but by eliminating the work that wasn’t adding value in the first place.
I don’t say this to be provocative. I say it because I’ve seen it from the inside. I’m living it. And if you’re a founder or CEO who’s been watching AI from the sidelines, wondering when it becomes relevant to how you actually operate — it’s now. It’s already here. The only question is whether you’re building with it or waiting for someone else to build it for you.
If you want a consulting partner who operates like this, now you know where to find one.