what a workflow automation consultant actually does in the first two weeks
search for a workflow automation consultant and most of what comes back is profiles, freelancer listings and articles asking whether you should just do it yourself. very little of it answers the question an owner actually has before hiring one: what will this person do, in what order, and what will i have at the end if i decide not to go further?
here's an honest version of that answer. it's the shape i use, and it's also a reasonable yardstick for judging anyone else's proposal. the short version: a good workflow automation consultant spends the first two weeks not building anything.
why the first two weeks shouldn't involve building
the most expensive automation mistake isn't a bug. it's automating the wrong step, or automating a step that should never have been automated at all. a consultant who starts by connecting tools on day one is guessing at where your time goes and what's safe to hand to software. the guess is usually wrong in the same direction: it assumes the problem is a missing tool, when the cost is usually in the gaps between tools you already have.
so the first stretch is diagnosis. it should be fixed in length, fixed in price, and end with something you own.
day 1: kickoff, about an hour
you walk the consultant through how work comes in and what happens to it. a good one mostly asks questions and resists proposing anything. if someone pitches a solution in the first meeting, they're pitching what they already sell, not what you need.
days 2 to 5: following the work
this is the part most proposals skip. the consultant follows a real piece of work end to end, a referral, an inquiry, an invoice, and records every point where a person retypes, re-checks, chases or re-decides something. rough hours go against each step so the expensive ones become obvious.
it also turns up the things nobody mentions because they've stopped noticing them. in one medical practice audit the systems themselves were fine. the cost sat between them: referrals logged and chased from memory, email doing the job of a coordination system, and the same physician's name spelled four different ways, so every report grouped by provider was quietly wrong. none of that shows up in a sales call. all of it shows up when someone follows the work.
days 6 to 9: what's allowed, and what's actually buildable
before anything gets priced, two checks. first, what your existing agreements and rules already permit: where data could move, what a vendor would need to sign, what has to stay where it is. often the answer is that you can build inside what you already have, which removes the slowest part of most projects. second, what's genuinely buildable against your current setup, as opposed to possible in principle.
this is also where every step gets sorted into three groups: automate fully, automate with a person checking, or leave alone. the third group matters most. anything involving professional judgment, a duty of care, or a consequence you can't undo stays with a person, and a good consultant says so plainly rather than quietly automating it because it was technically possible.
days 10 to 13: the plan gets written
the deliverable is a written plan meant for you to read, not a developer:
- a map of how the work actually moves today
- each step marked automate, automate with a checkpoint, or leave alone
- a phased build plan with a price against each phase, ordered so the earliest phase pays back first
the phasing matters. you should be able to stop after any phase and still be better off than when you started. if a plan only pays off once the whole thing is built, it's carrying more risk than it admits.
day 14: the walkthrough
about an hour. the consultant takes you through the plan, you push back, it gets adjusted. then it's yours, whatever you decide next: build it with them, build it with someone else, or do nothing for now with a clear picture of what the manual work costs.
what it costs, and how to judge a proposal
people searching for this usually want two numbers: what the first step costs, and whether it's worth it. mine is a fixed fee of $2,500 for the two weeks, credited in full toward a build if you go ahead. whatever anyone charges, these questions separate a real diagnosis from a sales process:
- is the first step fixed in length and price? an open-ended "discovery phase" tends to expand.
- will they follow real work, or just interview you? interviews capture what people think happens.
- will they tell you what not to automate? if every step comes back as automatable, the list was written to sell.
- do you keep the plan if you stop? you should, and you should be able to take it to someone else.
- will you own what gets built? accounts, data and rules in your name, not rented from the consultant.
there's a longer piece on judging a consultant when you can't evaluate the technical work yourself: how to choose an ai automation consultant.
where to start
if your work arrives faster than anyone can sort it, and your systems are fine but the gaps between them are manual, those two weeks are the place to start. the details, including what you receive at the end, are at /workflow-audit.