How to choose a technology consultant
Nine questions, and what a good answer sounds like.
In short
Judge a technology consultant on three things: whether they will build what they recommend, whether they will tell you when the answer is to do nothing, and what they leave behind when they go. Everything else follows from those. Ask who does the work day to day, what happens if the honest answer is a spreadsheet, who owns the code and the accounts, how the engagement ends, and what the first two weeks produce. A good answer is specific and slightly inconvenient for the person giving it. Ask for one thing you can open and use before you decide.
By Shaheer Shaikh, business operations consultant · Updated September 4, 2026
Work it out in a minute
Nine questions across revenue, capacity and control. Shows which of the three is costing you most, and where to start.
What we take on
Who does the work, day to day
Not who is on the call. A good answer names the person, says what else they are carrying, and does not change when the contract is signed.
Will you build it, or hand me a recommendation
Advice that never meets implementation drifts, because constraints only appear once somebody starts. A good answer is that they do both, and it changes what gets recommended.
What happens if the answer is to do nothing
Sometimes the tool you already pay for does this and nobody switched it on. A good answer includes a time they said so, and what they did instead.
Who owns the code and the accounts
A good answer is you, in your name, from the first day, with no dependency on them to change it later. Ask where the repository lives and whose card is on the hosting.
What do the first two weeks produce
A good answer is something you can read and act on, not a kickoff and a plan for a plan. Ask what you keep if you stop after it.
How does this end
Every engagement should have a shape and a finish. A good answer names a date, says what handover looks like, and does not require a retainer for the thing to keep working.
Show me one thing I can open
A tool, a page, a working system. Anything you can use for ten minutes tells you more about how somebody thinks than a case study written to look tidy.
What do you not do
A good answer is quick and specific. Somebody who does everything is either a much larger firm than they appear or has not decided yet.
The three things underneath all nine questions
The first is whether advice and delivery sit with the same people. A recommendation handed to somebody else to implement loses most of its value on the way, because the part that was hard was never written down. When the person who advises also builds, the advice has to survive contact with the work, and that changes what gets recommended before you ever see it.
The second is whether they will talk you out of something. Any consultant will find work in a business that has grown quickly. The useful ones tell you when the answer is a spreadsheet, or a setting nobody has switched on, or waiting six months. That costs them a project and it is the clearest signal you will get, so it is worth asking for a specific example rather than a promise.
The third is what is still standing after they leave. A system your team uses without thinking about it, documented well enough that whoever joins next year can change it. Ownership is part of this and it is concrete: the code in your repository, the accounts in your name, the credentials in your password manager. Ask on the first call, because it is much harder to ask on the last one.
Everything else is comparable across firms and mostly tells you about size rather than fit. Where they are based matters less than it used to. How many people they have matters only if the work needs them. What matters is whether one specific problem you have gets smaller, on a date, with somebody named against it.
Questions we get
Is a bigger firm safer?
Bigger buys you continuity and a bench, which matters on a long program with a lot of moving parts. It usually costs you seniority on the day to day work, because the people on the call are not the people doing it.
For one specific problem with a date on it, the question is who will actually be doing the work, and that question has the same answer regardless of firm size.
How do I compare proposals that look nothing alike?
Ask each one what the first two weeks produce and what you keep if you stop there. That question flattens the differences fast, because it asks for the same thing regardless of how the work is packaged.
Then ask what they would do if the budget were half. The answer tells you what they think the important part is.
Should I hire a consultant or an employee?
An employee is right when the work is a permanent function with a full week of load. A consultant is right when the work is a project, a decision, or a function you are not ready to hire for yet.
There is a longer version of this on the fractional page, which compares the three common shapes side by side.
What if we already know what we want built?
Then say so and ask for a build, not a discovery. A good consultant will still ask why, and will say if the thing you want is not the thing that helps, but they should not sell you a phase you do not need.
If you are sure, the useful check is a short review of the plan rather than a rerun of the decision.
How much of this applies to hiring an agency?
Most of it. The three underneath questions are the same: who does the work, will they talk you out of something, and what is left when they go.
The ownership question matters more with agencies, because more of them build on platforms and accounts that stay theirs. Ask early and get the answer in the proposal.
What is a reasonable first engagement?
Something with a date, one problem, and a deliverable you keep. Weeks rather than months, and small enough that being wrong about the firm is cheap.
If a first engagement cannot be described that way, that is worth asking about before it is worth worrying about.
More in the guides and every answer in one place.
Shaheer Shaikh, C-suite operations lead for a San Francisco AI company working with global pharma on clinical trial data and model benchmarking. Six Sigma on the process side, Anthropic certified on the Model Context Protocol, ten years across eight industries. See what we have built