This is the fastest single filter. A partner who has been through a few of these has a list of things they will not do: automate a decision nobody has made, promise an accuracy number before seeing the data, replace a process the client cannot describe, or send customer-facing replies without a human path.
If the answer is that everything you described sounds like a great fit, you are talking to someone whose incentive is the contract rather than the outcome. The best answer to some enquiries is a spreadsheet change or a setting in software you already own, and a partner who cannot ever reach that conclusion is not evaluating your case.
What happens when the model is wrong? You want thresholds, an exception queue, an audit trail and a person in the path — described specifically, not as a principle.
How will we know it is working? A metric agreed before the build, measured against a baseline you can see today. If accuracy is only mentioned after go-live, it will be a matter of opinion.
Where does our data go, and who can see it? Which provider, which region, whether inputs are retained, whether anything trains on your data. The answer should be immediate and precise.
What do we own afterwards? Code, prompts, configuration, infrastructure, credentials. Ask what happens on the day you leave — and whether it runs without them.
Who is actually doing the work? The people in the meeting, or a team you never meet. Both can be fine; being told the first and getting the second is not.
What is the price, and for what? A fixed price for a defined outcome is different from an estimate on a rate card. Ask what would make the price change and how that decision gets made.
Show something you built that failed. An honest answer exists in every real portfolio. A partner who has never had a project go sideways has either not built much or is not telling you about it.
How does it get maintained? APIs change, models are deprecated, volumes grow. Ask who watches it and what the arrangement costs after launch.
Can we start smaller? Any good partner will scope a first piece that proves the case. An all-or-nothing proposal usually means the value cannot be shown incrementally.
A proposal that names specific numbers with the source of each is a good sign — not because the numbers are large, but because someone measured something. A proposal full of ranges attributed to industry research is describing other companies. Ask which of the figures came from your own data, and watch what happens.
Another useful signal is how they talk about your existing systems. A partner who wants to replace the CRM, the helpdesk and the ERP in order to automate one process is solving a different problem than the one you have. The work should sit on top of what you run today unless there is a clear reason otherwise.
A freelancer is fast and cheap for a contained piece of work, and carries a real single-person risk on anything that needs to run for years. A large integrator brings process, compliance and staying power, at a price and a pace that suits a large programme rather than one process. A small specialist team sits between them: enough people that nothing depends on one person, small enough that the people who scoped the work are the ones who build it.
None of these is the right answer in general. What matters is the match to the size of what you are doing — and being honest about which one you are actually buying, since all three describe themselves in similar language.
We are the third kind, and there are cases where we say a freelancer or a large integrator is the better fit. We would rather say it in the first call than in month three.
Get the success criteria in writing, with the baseline measured before anything is built. Get the ownership terms in writing, including credentials and infrastructure. Get the exception behaviour in writing, because that is the part that quietly decides whether the system is usable.
Then start with one process. Not a platform, not a strategy engagement, not five workflows in parallel — one process, in production, measured against the baseline. Everything you need to know about a partner is visible by the end of that, and it costs a fraction of finding out later.
We use cookies for analytics — to see which pages bring enquiries. Nothing else, and nothing before you agree. Cookie Policy