A discovery is $2,500 and takes about two weeks. It ends with the quantified annual dollar value of your best workflow, the arithmetic behind that number, and the written acceptance test any software has to pass. Run from Orange County, California, remote across the US. If it doesn't surface 10x the fee, you don't pay.
Two weeks on one workflow. You end up with a dollar figure you can defend and a written test the software has to pass.
Two weeks, $2,500, one workflow. Usually the one bleeding the most money, mapped end to end with the people who actually run it.
You get the annual dollar value of fixing that workflow, with the arithmetic on the page so your CFO can argue with it. You get the acceptance test: the written conditions software has to meet before anyone gets to call it done. That's the product. No maturity model, no deck about your future.
Plenty of AI money gets committed before anyone checks the arithmetic. Pertama Partners' 2026 roundup found 42% of AI projects return zero ROI, and 61% of the projects approved on projected ROI were never measured after launch. A number nobody checks isn't really a number. Discovery exists to produce one you can check later, against reality.
Four documents. Plain language, no jargon tax.
All four are yours the day of the readout. See what a build looks like if the answer turns out to be software.
If discovery doesn't surface at least 10x the fee in quantified annual value, you don't pay.
Read that precisely, because it's precise. The threshold is $25,000 a year in value we can show the arithmetic for. Find it, and the fee stands. Don't find it, and there's no fee, and you keep the map anyway.
The guarantee covers the value, not the build. Say the two weeks turn up $400,000 of annual value and the honest recommendation is to fix the process rather than buy software. Guarantee met, fee stands. You paid $2,500 to not spend a build budget. If the value isn't there at 10x, that's our miss and we eat it.
Five steps, about six hours of your team's time in total.
Week one earns its keep on data alone. Gartner puts 85% of AI project failures on data quality, and IDC has 65% of companies citing data readiness as the blocker. Both are cheaper to discover in a mapping session than in month four of a build.
| Free vendor scoping call | Internal analysis | Portillo discovery | |
|---|---|---|---|
| What it costs | $0, plus four hours of your senior people | Salary you're already paying, plus the work that stops while it happens | $2,500 flat, waived if the 10x threshold isn't met |
| Who it serves | The vendor's pipeline. The scope fits what they sell. | The team writing it, and whichever department has the most pull | You. The recommendation can be to build nothing. |
| What you keep | A proposal written to be accepted | Whatever got written down, if anyone wrote it down | Map, valuation, acceptance test, recommendation. Take them anywhere. |
| How the number is verified | Vendor's estimate of your savings | Internal estimate, rarely audited after launch | Your hours and rates, arithmetic shown, signed off before the readout |
| What it produces next | A statement of work | A slide that circulates for a quarter | A test the software has to pass, or a documented reason to stop |
Nobody's being dishonest in column one. A free scope is a sales document, and it does its job well. It just isn't a deliverable, and you can't take it to another firm.
Because the free version can't say no. MIT's NANDA report found 95% of enterprise AI pilots deliver no measurable P&L return. Those pilots weren't scoped by fools. They were scoped by people whose next invoice depended on the answer being yes.
Paying for the analysis buys the option of a bad answer. It also buys documents that exist independently of us: the workflow map is true whether or not you ever hire anyone, and the acceptance test is a fair bar for any vendor you hand it to.
$2,500 is priced to be small next to the mistake it prevents. Build bands start at $5,000 for a narrow integration and clear $250,000 at the complex end. Two weeks of thinking against that is not a hard trade, and if the thinking doesn't pay for itself you don't pay for the thinking.
Builds are priced at 10 to 20% of the measured annual value, never by the hour. Hourly billing pays us for slow. The percentage only works because discovery produced a real number first.
| Work | Range |
|---|---|
| Custom web app or line-of-business system | MVP $15k–75k. Complex $250k+. |
| AI or LLM integration | Strategy or pilot $5k–25k. Mid-size $25k–100k. |
| RAG or knowledge copilot | Build $8k–45k, plus $650–1,750/month operations. |
| AI consulting retainer | From $2,500/month. |
| AIOps / automation retainer | $2k–15k/month. |
Every build carries the same promise: we work it until it passes your acceptance test. That promise is only safe to make because the test was written in week two, by both of us, before any money moved.
The engineers who map your workflow are the ones who build it. Nothing is thrown over a wall to a separate delivery group, and your account doesn't get handed to a junior once the paperwork is signed. We keep the number of live engagements deliberately small so that stays true, which sometimes means a start date six weeks out. We'll tell you that at the readout rather than after you've signed. What you get in exchange is in MIT's numbers: buying from a specialized vendor ships a working AI system about 67% of the time, roughly twice the rate of building it internally. Groovyweb's 2026 build-versus-buy breakdown puts an in-house AI team at $1M to $1.8M loaded in year one, with three to six months of ramp before it produces anything.
No, and the documents are built to make leaving easy. The map, the valuation and the acceptance test are written to be read by anyone. Hand them to another firm or to your own team. The acceptance test in particular is portable on purpose: it's the fairest way to hold any vendor to a result.
If you do build with us, the code is yours outright. It isn't resold to your competitors, it isn't licensed back to you, and there's no hosting arrangement designed to make leaving painful. You get the repository.
There's no retainer attached to discovery either, and no automatic anything. Two weeks, four documents, done. Whatever happens next is a separate decision made with better information. If you'd rather see how we work before spending anything, the AI Pilot Autopsy is free. Two pages, emailed back as a PDF, nothing to fill in and no sequence afterwards.
No. Discovery is priced on its own because it stands on its own. Crediting it would make it a deposit, and a deposit quietly pushes the answer toward build. The whole point is that don't build this stays a live option on the last page.
Roughly six hours across two weeks, spread over four or five people. The executive who owns the budget needs to be in two of those sessions. The people who actually run the workflow need to be in the mapping session, because they are the ones who know where it really stalls.
Yes. Most of it happens on video and works fine across the US. If you are in Orange County, the mapping session is better in person, and Irvine, Newport Beach, Costa Mesa, Santa Ana, Anaheim, Tustin and Huntington Beach are all a short drive.
That is the normal case, and it is worth knowing before you spend build money. Gartner puts 85% of AI project failures on data quality. The workflow map names the gaps it finds, and the acceptance test is written so the software has to pass with your real data, not a clean sample.
A short written agreement covering scope, confidentiality, and the 10x guarantee. Nothing recurring and nothing to cancel. Because the guarantee is in writing, the fee settles after the readout, once you have the documents and can see whether the number is there.
Email Portillo Technologies at [email protected] with the workflow that is costing you the most. If you have an AI pilot that already stalled, ask for the AI Pilot Autopsy first. It is free, it is two pages, and it will tell you whether a discovery is even the right next step.
One email with the workflow and roughly what it costs you. We'll tell you whether it's worth two weeks, and we'll say no if it isn't. If a pilot already stalled, start with the free autopsy instead.
Replies inside two business days, from Jason, not a sales queue.