Field Notes · Agentic QA Strategy

What the Big 4's Agentic Testing Playbook Gets Wrong for UK Mid-Market Teams

Deloitte, KPMG, PwC and EY are all heavily invested in agentic AI for testing, and their capabilities are genuinely serious. They are also built for global enterprises with eight-figure transformation budgets. For UK mid-market software teams and scale-ups, the Big 4 playbook is the wrong shape — not because it is bad, but because it was never designed for you. Here is why, and what to do instead.

13 June 20268 min read

TL;DR

The Big 4 agentic testing model is platform-led: capabilities slotted into large delivery platforms, governed centrally, delivered over long engagements. It works well for the buyers it was built for.
For UK mid-market teams it mismatches on three axes: procurement weight, time-to-value, and where the knowledge ends up.
Microsoft's UK data — more than half of organisations with no formal AI strategy — shows most of the market is not structurally ready to consume a platform engagement.
The alternative is not cheaper consultants. It is a different model: senior engineers, scoped to your application, no platform to procure, knowledge that stays with your team.

The model, fairly described

It is worth describing the Big 4 model accurately before arguing with it, because the capabilities are real. Deloitte has embedded UiPath Test Cloud into its Ascend delivery platform. KPMG sells quality engineering through G-Cloud with modular automation frameworks. EY has signalled an ambition to deploy more than a hundred thousand agents; PwC reports tens of thousands already running in client operations. These are not slideware.

The shape is consistent: a large engineering and delivery platform into which testing capability is slotted, governed centrally, and delivered through a long-running engagement. For a FTSE 100 buyer with the budget to absorb a six-month onboarding and the procurement function to manage a platform relationship, this is a rational way to buy. The scale of the buyer matches the scale of the model.

The problem is not the model. It is the assumption that the same model, shrunk, fits a buyer who is nothing like a FTSE 100.

Three structural mismatches

For a UK mid-market software team or a scale-up, the Big 4 playbook mismatches on three axes, and none of them is fixed by negotiating the rate.

Procurement weight. A platform engagement assumes a procurement function that can scope, govern, and manage a multi-month platform relationship. Most mid-market teams do not have that capacity to spare, and the overhead alone can exceed the value.
Time-to-value. The platform model delivers value after onboarding — measured in quarters. A scale-up needs agentic QA value against a specific application in weeks. The clock the platform runs on is the wrong clock.
Where the knowledge lands. In the platform model, the capability lives with the consultancy and its platform. When the engagement ends, so does much of the know-how. A mid-market team needs to own the capability, not rent it.

These are not complaints about quality. The Big 4 do serious work. They are mismatches of shape: the right tool for a different buyer.

The platform trap

The deepest mismatch is lock-in. A platform-led engagement ties your agentic QA capability to a platform you do not own and cannot easily leave. The frameworks are generic by design — they have to serve every client — so they encode little about your specific application, and what they do encode stays inside the platform.

For a global enterprise, that trade is acceptable; the platform relationship is one of many, and the governance scale justifies it. For a mid-market team, the same trade means your testing capability now depends on a vendor relationship priced for someone an order of magnitude larger than you, with recurring fees that assume a budget you do not have. You have solved a testing problem by acquiring a procurement liability.

Lock-in is a rational trade when you have the scale to manage the platform relationship. For a scale-up it is just a liability priced for someone else.

The hidden cost of generic frameworks

A standard delivery framework is the Big 4's efficiency engine — it lets them deploy consistent quality across hundreds of engagements. But a framework that must work for every client knows nothing about yours until someone teaches it, and that learning is billed.

Agentic QA value comes from understanding the specific application: its data shapes, its failure modes, the funnel that actually matters to the business. A generic framework starts from zero on all of that. The mid-market buyer ends up paying senior-consultancy rates for the early weeks of context-building that a smaller, application-first team would have started with on day one. The framework is an asset for the consultancy and a cost for you.

What mid-market teams actually need

Strip away the platform and the question gets simple. A UK mid-market team needs agentic QA value against its specific application, quickly, delivered by people senior enough to make the right calls, structured so the team owns the capability at the end. Microsoft's UK Cloud data — more than half of UK organisations with no formal AI strategy — underlines that most of the market is not structurally ready to consume a platform engagement, and does not need to be.

This is the same point we make about getting an agentic QA pilot to production: the thing that creates value is mostly senior discipline applied to your context, not a platform you plug into. The platform is the Big 4's answer to scale. For a team that is not at that scale, it is overhead pretending to be capability.

A leaner alternative

The alternative to the platform model is not cheaper consultants doing the same thing for less. It is a different model entirely: senior UK engineers working hands-on alongside your team, scoped to your actual application, CI topology, and business context rather than to a platform's standard framework.

There is no platform to procure, no multi-month ramp before value appears, and no junior-led delivery dressed up by a senior sales team. The work is structured so the knowledge transfers to your team during the engagement, so the capability stays with you when we step back. It is, deliberately, the opposite of the platform play — context-first, senior-led, and built to make you independent rather than dependent. That is the model GVK Technologies runs, and it exists precisely because the Big 4 playbook leaves the mid-market underserved.

The Big 4 sell you a platform because platforms scale to their business. A mid-market team does not need a platform — it needs senior people who understand its application and leave the capability behind.

Key takeaways

The Big 4 agentic testing model is platform-led and genuinely capable — for the eight-figure-budget enterprises it was built to serve.
It mismatches mid-market teams on three axes: procurement weight, time-to-value, and where the knowledge ends up.
Platform lock-in is a reasonable trade at enterprise scale and a pure liability for a scale-up priced as if it were one.
Generic frameworks bill you for context-building that an application-first team starts with on day one.
The alternative is senior, context-first, no-platform delivery that leaves the capability with your team — not cheaper versions of the same engagement.

FAQs

Are you saying the Big 4 agentic testing capabilities are bad?+

No. They are serious capabilities, well suited to the global enterprises they were built for. The argument is about fit, not quality. A platform-led model designed for an eight-figure transformation budget is the wrong shape for a mid-market team, however good the underlying work is.

We're a scale-up but growing fast. Shouldn't we adopt the enterprise model early?+

Adopting an enterprise-scale model before you are at enterprise scale usually means paying for overhead you cannot yet use. The capability that helps you now is senior, application-first delivery that leaves know-how with your team. If and when you reach the scale where a platform earns its keep, you will be in a far stronger position to evaluate one — because you will own the capability rather than rent it.

Doesn't a platform give us more governance and consistency?+

It gives the consultancy consistency across many clients. For a single mid-market team, governance and consistency come from senior people applying good discipline to your specific context, integrated with your existing CI and controls. You do not need to acquire a platform relationship to get a well-governed test suite.

What happens to the capability when the engagement ends?+

That is the central difference. In the platform model, much of the capability leaves with the consultancy. In the model we run, knowledge transfer is built into the engagement, so your team owns and can extend the agentic QA capability after we step back. Independence is the goal, not a follow-on contract.

How fast can a leaner model actually deliver value?+

Because there is no platform to procure and no multi-month onboarding, work starts against your real application immediately, and value is measured in weeks rather than quarters. The exact timeline depends on your application and CI, which is what a short scoping conversation establishes.

Need agentic QA value without the platform?

We scope context-first agentic QA engagements for UK mid-market teams and scale-ups: senior engineers, your application, your CI, weeks not quarters, and no platform to procure. The capability stays with your team when we step back.

Scope a context-first engagement

About the authorVenkata Kari · Founder, GVK Technologies

Twenty years in QA leadership, much of it spent on the side of the buyer trying to make enterprise delivery models fit teams they were never designed for. GVK Technologies is the deliberate alternative: senior engineers, scoped to your application, no platform to procure, capability that stays with your team.

All posts