Staff Augmentation Is Dead. It Just Doesn't Know It Yet.

I’m going to say something uncomfortable: the business model that’s sustained thousands of engineering consultancies for the last two decades is dying. Fast.

I’m Rich Hay, co-founder of bigspark — a small-to-medium sized engineering consultancy that’s spent 7 years building software, data platforms, and cloud infrastructure for some of the UK’s biggest enterprises. We’re good at what we do. Our clients trust us. Our engineers are sharp.

And none of that will save us if we don’t change.

What I’m seeing

I spend a lot of time inside talking to people, in large enterprises, in small businesses, to my friends who work in other industries. What I’m watching happen right now is genuinely unprecedented. Agentic AI programmes are being rolled out across entire organisations — not just engineering, but operations, finance, compliance, customer service, HR. Every function. Every level.

In engineering, the model is shifting from “humans write code, AI assists” to “AI writes code, humans groom the context.” But it’s not stopping there. In compliance, AI agents are reviewing regulatory filings. In operations, they’re managing workflows end-to-end. In finance, they’re doing analysis that used to take teams of analysts weeks.

This isn’t happening in 5 years. It’s happening in 12-18 months. Maybe faster. Speed to market is everything, adaptability is key.

When a large enterprise is actively planning for a world where they need a fraction of the headcount across the board — and they’re not being pessimistic, they’re being practical — you have to ask yourself: where does that leave a company that sells people’s time?

The Magnificent 7 problem

Microsoft, Google, Amazon, Meta, Apple, Nvidia, Anthropic and OpenAI are pouring hundreds of billions into AI. They’re not building better developer tools. They’re building human replacements.

Copilot writes code. Cursor builds features. Codex runs autonomously. Every month, the bar for “what requires a human” gets higher. Every month, the economic argument for hiring a team of contractors gets weaker.

If you’re running an engineering consultancy and your strategy is “we’ll keep doing what we’ve been doing but mention AI on the website,” you’re already dead. You just haven’t noticed yet.

The honest bit

Here’s where I get personal. bigspark’s earnings have been reflective of your average engineering led consultancy — great growth in the early years, a bumpy ride in the awkward teenage years, things are looking bright right now but we can see the future fast approaching. The staff augmentation market — which was always a big chunk of our revenue — is contracting. Not because we’re doing bad work. Because the market itself is evaporating.

I could pretend this isn’t happening. I could wait for the board meeting, commission a strategy review, hire a consultant to tell me what I already know. But I don’t have that luxury. None of us do.

So I’ve been doing something about it. Nights. Weekends. Bank holidays. While keeping the day job running. Because the time to act isn’t next week/month/year — it’s now.

What we’re actually doing

We’re transforming bigspark into an AI-native business. Not “we use AI tools” — every company says that. I mean we’re building our own internal platforms and tooling for creating AI-powered workflows and agent interfaces.

The idea is simple but powerful: a configurable platform where you define a persona — give it domain knowledge, connect it to data sources, equip it with tools — and it becomes a specialist AI agent for that job. We’ve already built our first one (a fraud investigation agent that synthesises years of domain expertise into an interactive workflow). It’s genuinely impressive, and it took a fraction of the time traditional development would. Hours and days not weeks.

But here’s the real vision: I don’t just want engineers using this. I want our head of finance building their own workflows. I want ops people creating their own tools. I want the platform to democratise the ability to build useful AI-powered things — without needing a 6-week sprint cycle and a team of developers.

And eventually? I want to bring this capability to our clients. Not as consultants telling them what to do — as practitioners who’ve already done it to ourselves.

The three-step plan

Step 1 (now): Transform internally. Get everyone AI-enabled. Build more agent personas. Prove it works on ourselves.

Step 2 (3-6 months): Use the platform to deliver client work faster and better. New service lines: AI strategy, agent development, enablement.

Step 3 (6-12 months): Productise the platform. Let other businesses build their own AI workflows using our tooling.

Why I’m writing this publicly

Because I think the industry needs more honesty and less theatre.

Too many LinkedIn posts about AI are either breathless hype (“AI will solve everything!”) or vague corporate positioning (“we’re excited to announce our AI practice”). Neither is useful.

I’m going to document this journey — warts and all. What works, what doesn’t. What’s hard, what’s surprisingly easy. The real numbers, the real failures, the real wins.

At least one article a week from me and others to follow from the core team. The unfiltered story of turning a traditional engineering consultancy into an AI-powered business in real time.

If that sounds interesting, follow along. If you’re going through something similar, I’d love to hear from you. And if you think I’m wrong about staff augmentation — tell me. I’d genuinely love to be wrong about this one.

But I don’t think I am.

This is article 1 in a series documenting my company’s transformation into an AI-native business. Next week: Gathering the team to think about tooling, ways-of-working, areas of focus for the next week and reflect on what we’ve done in the last week.