◆ STRATEGY · 8 MIN READ ◆

Build custom AI, or buy off the shelf? Answered honestly.

Every vendor says buy. Every agency says build. Here's the framework we give clients even when it costs us the project.

Here's a sentence you'll rarely hear from an AI services company: most businesses should buy most of their AI. Off-the-shelf tools are excellent, cheap, and improving monthly. The interesting question is identifying the specific places where they're not enough — because that's where custom AI stops being an expense and becomes an asset.

01

The three questions that actually matter

1. How unique is the workflow? If a thousand other companies do this task the same way (meeting notes, generic email drafting), someone has already productised it beautifully. If the workflow encodes how you specifically win — your qualification logic, your document standards, your pricing judgement — no product will fit without painful compromise.

2. How sensitive is the data? Client files, financials, health records, anything regulated: you need control over where data flows, how it's retained, and who can see it. Custom builds let you draw those lines exactly; SaaS tools draw them for you.

3. Is this capability strategic? Would a competitor with the same tool erase your advantage? If your edge lives in the workflow, renting the same software as everyone else means renting away the edge.

02

When buying is obviously right

Buy without guilt when the job is generic, the data is low-sensitivity, and speed matters: transcription, grammar and writing aids, generic chat assistants, standard analytics, scheduling. The market has invested hundreds of millions perfecting these. Your £30–£100 per month buys the output of engineering teams you could never hire.

The 80% rule: if an existing tool covers 80%+ of your need and the missing 20% is inconvenience rather than deal-breaker — buy. Configuration beats construction.
03

When building pays for itself

Custom development earns its keep in three situations we see repeatedly:

The workflow IS the business. A visa consultancy's document-checking logic, a lender's risk triage, a clinic's intake pathway. Automating your secret sauce with your own system compounds the advantage instead of leaking it.

Integration is the value. The win isn't any single AI feature — it's AI stitched through your CRM, your documents, your approval chain. Off-the-shelf tools live beside your systems; custom AI lives inside them.

Unit economics at scale. Per-seat SaaS pricing that looks cute at 5 users becomes a six-figure line item at 200. Past a threshold, owning the system beats renting it — permanently.

FIG. 01 — THREE-YEAR COST CURVESILLUSTRATIVE · 60-SEAT TEAM
Illustrative only: rented software compounds with seats and years; built systems front-load cost, then flatten. The crossover point is where strategy lives.
"Buy the capabilities everyone needs. Build the capability that makes you, you."
04

The total-cost illusion

Cost people compareCost that actually decides
SaaS: monthly feeFee × seats × years + workflow compromises + data exposure + switching pain
Custom: build quoteBuild + maintenance — minus hours saved, errors avoided, and the asset you now own

Both sides hide costs. SaaS hides them in per-seat multiplication and lock-in; custom hides them in maintenance — every built system needs an owner, monitoring, and occasional updates as models improve. An honest comparison prices three years of each path, not month one.

Red flag: any vendor — SaaS or agency — who answers "build or buy?" before asking about your workflow, data, and scale is selling, not advising.
05

The hybrid default

In practice, the winning pattern is rarely pure: buy the commodity layers, build the differentiating core. Use excellent off-the-shelf tools for transcription, writing, and generic assistance — and invest custom development in the one or two workflows where your business logic lives. That's how you get 90% of the AI advantage at 30% of the naïve "build everything" budget.

The build-vs-buy question, answered honestly, is almost never either/or. It's a sorting exercise — and the sorting criterion is simple: how much of "you" is in the task?

Build or buy? Get your steer.

Three quick calls — we'll point you at the likely answer.

1. Is this workflow how you specifically win (your secret sauce)?
Yes, it's differentiating
No, it's generic
2. How sensitive is the data involved?
Highly sensitive
Low sensitivity
3. Scale you're operating at?
Large / many users
Small team

Answer all three for your recommendation ☝️

Want a straight answer for your case?

Tell us the workflow. We'll say build, buy, or hybrid — and if the answer is 'buy a £30/month tool', we'll tell you that too.

Book a free discovery call ✦