Free guide15 min read

Build vs Buy: AI Agents for Enterprise

A decision framework for CTOs and Heads of AI in 2026.

By Daniel Usvyat · Founder & Principal, USQRD

Skip the read — get a recommendation in 2 minutes

7 questions. One honest answer: Build, Buy, Partner, or Wait.

Why this guide exists

Most enterprise AI agent conversations end in one of two places: a £2M internal build that ships nothing in year one, or a vendor licence that doesn't fit the data model. Both are avoidable. The decision you're actually making isn't whether to deploy agents — it's which path gets you to production fastest at acceptable risk and cost.

This is a framework for making that decision honestly. It's not a sales document. If the answer for your situation is “buy a tool,” I'll say that clearly.

The four paths

Every enterprise considering AI agents is choosing between four paths, whether they frame it that way or not.

1.Build in-house. Hire a Head of AI, build a team of 4–8 engineers, give them 12–18 months. High control, high cost, long timeline.
2.Buy off-the-shelf. License an existing agent platform (Glean, Writer, Cohere, etc). Fast to deploy for standard use cases, but you're on their roadmap.
3.Partner. Bring in a specialist delivery firm to design, build, and ship the system. Fixed scope, senior people, full handover. 8–12 weeks to production.
4.Wait. Stay on ChatGPT Enterprise or Claude for Work. Re-evaluate in 6–12 months.

Each is the right answer for some organisations. The wrong answer for most.

The decision framework — seven questions

Answer these honestly. They map to the four paths above.

1. Do you already have a Head of AI in seat?

No → rule out Build until you do. Skipping this step is how £2M disappears into unfinished prototypes.
Yes → continue.

2. Is AI a core differentiator for your business, or an efficiency play?

Core differentiator (AI is reshaping your product or your competitive moat) → Build or Partner.
Efficiency play (internal productivity, document processing, support) → Buy is likely sufficient.

3. What's your realistic time-to-value?

12+ months acceptable → Build is viable.
6 months → Partner.
6 weeks → Buy.

4. How sensitive is the data?

Highly regulated (financial, health, government, sovereign) → rules out most off-the-shelf platforms. Partner with a firm that deploys into your own infrastructure.
Standard enterprise → all four paths are open.

5. What's your AI hiring budget — including the 50% you'll lose to attrition in year two?

£1.5M+ → Build is realistic.
£500K–1.5M → Partner.
<£500K → Buy or Wait.

6. Do you have a clear, scoped use case you're confident in?

Yes, specific and narrow → Partner can ship it in 12 weeks.
No, still exploring → Buy a tool to learn cheaply first, or bring in a Fractional Head of AI to scope it before spending delivery budget.

7. What's your appetite for delivery risk?

Low → Buy or Partner (with fixed-scope contracts only).
High → Build, but only with a strong Head of AI already in seat.

What each path actually costs (12-month, all-in)

Numbers are for a single mid-complexity production agent deployment.

Build in-house

Head of AI: £180–250K loaded
4 engineers × £130K loaded: £520K
Tooling, infrastructure, model spend: £80–150K
Recruiter fees (assume one hire fails): £40K

Total: £820K–£960K. Time to first production agent: 9–14 months.

Buy off-the-shelf

Platform licence (50–500 seats): £40–200K
Implementation partner: £50–150K
Internal team time (1 PM, 1 engineer, 0.5 FTE business owner): £200K loaded

Total: £290–550K. Time to first use case: 4–12 weeks.

Caveat: you're on the vendor's roadmap permanently.

Partner

Fixed-price production agent: £80–180K
Fractional Head of AI (retained): £60–120K over 6 months
Internal team time (1 PM, 1 engineer): £150K loaded

Total: £290–450K. Time to first production agent: 8–14 weeks.

Wait

ChatGPT Enterprise / Claude for Work: £20–100K depending on seats
Internal experimentation: £100K

Total: £120–200K.

Risk: competitors who chose Build or Partner are 12 months ahead by next year.

Common failure modes

You've chosen wrong if any of these are happening six months in:

You built — and:

Your Head of AI is doing engineering, not leading
Your engineers are arguing about which framework to use, not shipping
You have three half-finished prototypes and nothing in production
Procurement is blocking your model spend

You bought — and:

The platform doesn't fit your data model and the vendor's roadmap won't help
Your team is building wrappers around the vendor's API to make it usable
Your competitors have the same tool and are getting the same 80%

You partnered — and:

The partner won't show you their code
They're billing time-and-materials with no fixed scope
The senior people in the pitch aren't the people delivering

You waited — and:

A competitor announced something you can't match in under 12 months
Your team has lost one or two senior engineers to AI-native employers
Your board is asking quarterly questions you can't answer

When each path is actually right

Build in-house if:

AI is your core product moat, not a feature
You can hire a Head of AI in the top 5% (and pay for it)
Your timeline tolerates 12+ months to first production system
Your data and regulatory environment make external partners impossible

Buy if:

The use case is well-served by an existing platform
Speed matters more than differentiation
You're solving an efficiency problem, not a competitive one
You can live with vendor lock-in at the current price trajectory

Partner if:

You have a clear use case but no senior AI capability internally
You need production-grade delivery in under 16 weeks
You want to build internal capability through the engagement, not from scratch afterwards
You'd rather pay for outcomes than headcount

Wait if:

You don't yet have a use case where the ROI is obvious
Your team is consumed by other priorities for the next six months
The space is moving so fast in your vertical that a 6-month pause is defensible

What we do

USQRD is a senior-only AI delivery studio. We work with organisations in the Partner lane.

Fixed-scope, fixed-price engagements:

→30-day AI Roadmap — for organisations that need to make the build-vs-buy decision with rigour before spending delivery budget. Delivered by a senior practitioner. Ends with a recommendation you can put in front of your board.
→12-week Production Agent — for organisations that have decided and want to ship. Fixed price, your infrastructure, full team handover.
→Fractional Head of AI — for organisations that need senior AI leadership without the full-time hire. Monthly retainer, exit at any time.

Claude Code-first. Senior-only. No pyramid, no offshore handoffs.

If that sounds like what you're looking for, the next step is a 30-minute discovery call — no deck, no pitch, just an honest read on whether it's worth pursuing.

Get the full guide as a PDF

Drop your email for the printable version, plus occasional field notes from production AI engagements. Or download the PDF directly.

Already know which path you're on?

Book a discovery call

30 minutes with a senior practitioner. No pitch — just an honest read on whether Partner is right for you.

Book a discovery call

Written by Daniel Usvyat, founder of USQRD. Daniel has led AI and engineering functions at The Guardian, Anglo American, Sky, and multiple YC-backed startups. He's been building production AI systems since 2015.