Enterprise AI Platforms: who sells what, and who carries the risk
If you sit on a board, or run operations or transformation in an enterprise, a government entity or a segment-leading company, and you're evaluating how to put AI into your operations, you're caught in a paradox: every vendor sells a category — a platform, a strategy, engineering hours — and none of them sells the operation moved.
Palantir AIP is extraordinarily powerful but priced for the Fortune 500. McKinsey will build you a beautiful AI strategy but charge $2M before a line of code is written. Databricks will give you a world-class data platform but expects you to hire the data engineers to use it. And the IT services firms — EPAM, Thoughtworks, Globant — will sell you engineering hours at $200/hour while quietly worrying that AI is about to make their business model obsolete.
This guide evaluates the realistic options for organisations that want one operation moved, with a number to show for it, without an 18-month timeline.
Last updated: September 2026
Every company is choosing between five paths.
Understanding which path you're on is more important than evaluating individual vendors. Each has genuine strengths — and a structural gap that matters if what you need is an operation moved, not a category bought.
We find the one operation capping growth, measure what it costs today, and fix it until the number moves. That is the category Xamun occupies — the operation moved, not the platform, the deck or the hours — and it closes the gap in all four paths above.
Discovery names the bottleneck and the number. Build and launch puts the software live, typically within weeks, instrumented to that number from day one. Done is when the number holds unattended — and for three years after, if it slips, we re-engineer it free, whatever the cause. Two Minds and one graph memory, under the hood. That is the structural difference between a project and an operation that stays moved.
The real numbers. No euphemisms.
Based on an $80M-revenue healthcare services company implementing AI across patient scheduling, clinical workflow automation, and compliance reporting.
| Dimension | Palantir AIP | McKinsey / BCG | EPAM / TW | Internal Build | Xamun |
|---|---|---|---|---|---|
| Annual cost | $1M–$2M/yr | $2M–$4M (strategy + build) |
$400K–$800K | $800K–$1.5M (salaries only) |
Licence or per transaction per transaction from go-live, or one capital fee — no hours, no seats |
| Time to first value | 3–6 months (intelligence only) |
18 months (6 mo strategy + 12 mo build) |
8–14 months | 18 months (6 mo hiring + 12 mo building) |
Weeks, typically a short Discovery, then one live operation |
| Internal team required | Data engineers + platform admins |
PM to manage consultants |
Product owner + architect |
Full AI team: 3–5 engineers |
Business sponsor No technical staff needed |
| Ongoing intelligence | Yes — core strength | No — ends with engagement | No | Only if you build it | Under the hood — not sold as a subscription |
| Software delivery | No — platform only | No — strategy only | Yes — core strength | Yes — if you can hire | Yes — built, launched and held |
| Outcome governance | Partial — data-level | No — project ends | No — team rolls off | Only if you build it | Yes — one agreed number, held three years |
Ranges are illustrative, based on publicly available pricing and industry benchmarks. Actual costs vary by scope and complexity. We've been honest about where competitors excel — Palantir's intelligence capability is genuinely world-class, and EPAM's engineering discipline is top-tier.
What happens after the first project?
Most AI transformation evaluations focus on "which vendor is best?" The better question is: what happens when the first engagement ends?
Which path fits your situation?
Honest recommendations based on what you are actually buying and the bench you have to run it. Not every organisation should choose Xamun.
You want a general-purpose data and decision platform across many domains, and you have the internal capability to deploy, build on and maintain it.
You know the operation that is capping growth, or want it found. You need the people who name it to build it and hold the number — not a strategy vendor and an implementation vendor with a handoff between them.
Sometimes the bottleneck is knowing which operation it is. Discovery starts with a constraint inventory — what is countable, constrained and reversible, ranked — before you commit budget.
Questions we hear from CIOs.
It depends on what you are buying. If you want to own and operate a general-purpose platform and have the engineering bench to build on it, Palantir AIP is genuinely world-class at creating a digital model of your business. McKinsey sells strategy but not software. EPAM and Thoughtworks sell engineering hours but don’t tell you what to build. If what you need is one operation moved — the bottleneck named, the number measured, the software built and held until the number holds — with the vendor carrying the delivery risk, that is what Xamun sells, to boards, CIOs, COOs and transformation leads in enterprise, government and segment-leading companies. The honest answer is that most “enterprise AI platforms” sell a category; Xamun sells the operation moved.
The range is enormous. A McKinsey strategy engagement plus an EPAM build runs $2M–$4M over 18 months. Palantir AIP is $1M–$2M per year. Building an internal team costs $600K–$1.2M in salaries before they deliver anything. Xamun has no annual cost of that kind and no rate card. You buy it the way it suits your business: per transaction — a small fee on the unit your own business is paid on, from go-live, no upfront CAPEX — or a licence — one capital fee for the bottleneck, full source delivered under a source-available licence, final payment at done. Either is a fraction of the alternatives for the $80M healthcare company in the table above. The difference isn’t because Xamun does less; it’s because we sell the number moving, not the hours spent moving it.
Not entirely, and anyone who claims otherwise is selling something. The judgment, relationship management, and organisational change expertise that top consultants bring is genuinely valuable. What AI can replace is the data gathering, analysis, and report generation that accounts for 60–70% of consulting engagement time. Xamun’s Discovery, with Xamun Intelligence doing the reading under the hood, ends in the bottleneck named, the number measured and how you buy it agreed — work that would take a consulting team 6–12 weeks and $500K+. The inventory of operations is a by-product, and free. If you then need deep organisational change management, a focused consulting engagement on top of a moved operation makes sense.
With a traditional consulting-plus-build approach: 18 months (6 months for strategy, 12 months for implementation). With Palantir: 3–6 months for the intelligence platform, but you still need someone to build operational software on top. With Xamun: a short Discovery names the bottleneck and the number, and the operation is typically live within weeks, instrumented to that number from day one. The structural reason for the speed is that the people who name the bottleneck build the software — there is no gap between “strategy” and “requirements” where months typically disappear.
Start with one bottleneck.
A short Discovery names the bottleneck, measures the number, and agrees how you would buy it — and you’ll meet the people who’ll be inside your operation. Not sure which operation it is? Discovery starts with a constraint inventory. In-person or online.