DAY 1 · 90 MIN
We listen
A working session with the people who do the work daily, not the people who describe it in meetings. We ask what annoys them, where they wait, and what they've built workarounds for.
You bring: 2–4 practitioners
Most AI proposals ask you to buy a plan before anyone has looked at how your work actually flows. We'd rather look first, build something narrow against your real data, and tell you what we find — including when the answer is don't build it.
The hard part of an agent is never the model. It's the eleventh edge case in your approval chain, the field your team fills in inconsistently, the system whose API was last touched in 2019. None of that appears in a requirements document, because nobody thinks to mention it. It appears when an engineer sits with the work.
So we do the sitting first, before there's money on the table. It costs us five engineer-days. It saves you a six-figure decision made on assumptions.
Your people are needed on days one, two and five. Days three and four are ours.
Swipe →
A working session with the people who do the work daily, not the people who describe it in meetings. We ask what annoys them, where they wait, and what they've built workarounds for.
Every handoff, every system, every point where a person copies data from one screen to another. We draw it in front of you and you correct us until it's right.
Most of what you need, we've built before. We identify which of our twenty-five agents apply, what has to be built new, and where your systems will fight us.
A narrow working demo against a sandbox of your real data. Not a mock-up, not a video. Something your team can poke at and try to break.
Demo, findings, effort estimate and a recommendation written plainly — build this, build it later, or don't build it. Then the engagement is over unless you want more.
No watermark, no "confidential to IntegrateX," no clause preventing you from taking our workflow map to another vendor for a competing quote. If our estimate isn't the best one, you should know that.
We've had clients do exactly this. Some came back.
Your process as it actually runs, with automation points marked and the manual steps that should stay manual noted as such.
Running against your sandbox, with its limits stated. We tell you what it can't do yet as clearly as what it can.
The systems that will be difficult, the data that isn't as clean as anyone thinks, the approvals you'll need from IT.
Ranged, not falsely precise, with the assumptions written down so you can challenge them.
One page. Build, defer, or don't — and why. This is the part most vendors leave out.
If your security posture won't allow sandbox access inside five days — and in many enterprises it won't — we run the Sprint against a synthetic dataset you provide the shape of. It's a weaker demo, and we'll say so, but the workflow map and the estimate are unaffected.
There isn't a hidden one, but there is an honest one: this is how we sell. We think our chance of winning good work goes up when a client has watched our engineers think, rather than watched our salespeople present. Five days is what that costs us, and we'd rather spend it than a marketing budget.
The second thing worth saying plainly: we run a limited number of Sprints each month because they use senior engineers, not a pre-sales team. That's a real constraint, not a scarcity tactic. If we're full, we'll tell you when the next slot is rather than staffing it with someone junior.
And if the honest answer at the end is that you shouldn't build an agent for this, we say that. It happens often enough that we'd be lying to call it rare.
Remote by default, which is what makes five days practical. We run Sprints across European, Gulf and US time zones from our Ahmedabad, Dubai and California teams. On site is possible for the first and last sessions if you're near one of those offices, or if the engagement warrants travel.
An engineer who builds agents and a solution lead who has run integrations into systems like yours. Not an account manager with an engineer on standby. The person who maps your workflow is the person who would build it.
One, properly, beats three superficially. If you have several candidates, day one includes a short triage and we'll recommend which to take deep — that recommendation is itself useful, and the other workflows get noted in the findings pack for later.
Then we quote from the findings, and the demo we built becomes the starting point rather than throwaway work. Two paths from there: a fixed-price build we hand over complete, or a dedicated engineer on retainer working inside your team. Neither requires you to decide during the Sprint.
Bring the one people complain about. That's not a joke — sustained internal complaint is the most reliable signal of a workflow with real friction and real volume. If you'd rather talk it through first, a twenty-minute call before the Sprint costs nothing either.
One paragraph is enough to start. We reply within a business day with either a proposed week or an honest reason we're not the right fit.
Limited slots each month · Senior engineers only