A $7.7 billion deal just put the AI-consulting business model on trial.
In September, Sequence Holdings and Michael Dell's family office agreed to take Baldwin, an insurance broker, private for $7.7 billion — the largest AI-driven take-private to date. Sequence's business is buying companies and installing its own engineers inside them, permanently. The CEO, Michael Lee, explained the diagnosis on the No Priors podcast (Sep 24, 2026):
"Consultants are paid to stay, and software is sold to the workflow as it exists today."
That sentence is about a bank he bought. But it's also the sharpest question you can ask about the AI consultant you're about to hire — because the same incentive applies at every budget level. Services firms optimize to keep you as a customer, not to reinvent your business. The engagement-structure problem in AI consulting isn't a theory; it's what the people funding these deals say out loud.
The Dependency Machine
The pattern is subtle because every step feels reasonable on its own:
- The retainer that never ends. Discovery, roadmap, "ongoing optimization." Each renewal is smaller than the project of replacing them.
- The tool stack you don't own. Licenses in the consultant's accounts, workflows in their automation platform, playbooks in their Notion. The work lives inside their walls, on a meter.
- The knowledge that never transfers. They tune the prompts, they know why the triage rules route the way they do, they hold the context. You get the output, not the capability.
None of this requires malice. A services business grows by keeping clients; a client who has graduated doesn't pay a retainer. The incentives are the contract.
The Antidote: Buy Outcomes, Exit Ramps Included
You can use consultants well. The trick is to hire for a defined outcome with a built-in ending.
Structure for handover from day one. Before signing, agree on what "done" looks like and who owns it after: documentation you keep, training your team receives, tooling you can maintain without the advisor on the line. If the proposal doesn't contain a handover plan, ask for one. If they resist — that's your answer.
Prefer fixed-scope for implementation; keep retainers for true operations. "Build and wire up our invoice-exception automation for $X, including a runbook and one training session" is a project. "Manage your AI" is a subscription to dependency. Fixed-scope puts the incentive where you want it: finish, hand over, leave.
Own the accounts. Your API keys, your subscriptions, your platforms — set up under your business email, even if the consultant configures everything. If it's in their account, it's not your capability; it's their leverage.
Treat the runbook as the deliverable. The documentation of how the automation works — inputs, decision rules, exception paths, who's paged when it breaks — is the actual product. The workflow is just where the runbook runs.
Watch for the "celebrated persona" test. Lee made a related point worth borrowing: a company's celebrated persona decides who it can hire. At Blackstone, it's the investor; at his firm, it's the engineer. A consulting firm whose stars are salespeople will staff your account with handoffs. One whose stars are engineers embeds capability. Ask who does the work — and whether that person's job is to become unnecessary.
The Two-Sided Takeaway
There's a larger pattern here, and it runs in both directions:
- If you run a boring-but-profitable business — the plumbing firm, the clinic, the brokerage — AI-enabled acquirers like Sequence are becoming a realistic buyer class. One thing they pay for, beyond cash flow, is a business whose data is clean and whose workflows are documented. Every hour spent writing the runbook you'd need to fire a consultant is the same hour that raises your exit value.
- If you hire advisors, the test is simple and worth writing down: a good advisor makes themselves unnecessary on a schedule, and the contract is built to let them.
The counterpoint that pays for itself: embedded engineers at a holding company took BankSouth's commercial loan process from 30 days to 11, with 94% less consumer underwriting, in six months — because the engineers live inside the business, not outside it. The lesson isn't "never hire outside help." It's that transformation compounds when capability ends up inside your walls, and rents when it doesn't.
The Three-Question Contract Review
Before your next AI engagement, ask these in writing:
- What do I own on the day you're done? (Accounts, data, code, docs, access — in your names, exportable today.)
- What does my team know on the day you're done? (Training delivered, runbook handed over, at least one person who can change the logic without a call.)
- How much does it cost to keep you forever — and how much to replace you? (If staying is cheap and replacing is expensive, you're not buying capability. You're renting your own processes.)
Bottom Line
The AI consulting wave has an incentive problem the customers rarely see: the vendor's best month is the one where you never graduate. Structure for the opposite. Fix the outcome, fix the ending, own the accounts, and pay happily for the handover — because the runbook is the only deliverable that keeps working after the invoice stops.
If you want the full framework for building AI operations you own — not rent — we wrote it down: The AI Agent Owner's Playbook — $49 CAD, instant download. It's the owner's-manual version of the same thesis: capability belongs inside your business, on your accounts, in your documentation. And we're currently running a public 14-day sprint with daily revenue logs — $0 and counting, reported honestly — while we apply it to our own store.
This piece draws on Michael Lee's Sep 24, 2026 No Priors interview (Sarah Guo), Axios coverage of the Baldwin take-private (Sep 14, 2026), and Sequence's published BankSouth figures. Illustrative patterns (retainer structures, account ownership) are generalized from common engagement models, not any specific client relationship.
Top comments (0)