3 Engineers, 20-Person Output: Math

A 20-person consulting engagement costs £150-280k/mo. A 3-person AI-augmented pod costs £20-40k/mo and ships comparable scope. Not headcount math.

Anystack Engineering

The Big-Five math doesn't survive AI

For the last twenty years the dominant enterprise consulting model has been the same: a 15–25 person team, led by one or two senior architects, with the work distributed across mid-level engineers, junior delivery staff, project managers, and a partner who shows up for steering committees. Day rates run £800–1,400 for the seniors, £400–700 for the mid-levels, and £200–400 for the juniors. A six-month engagement: £1.5m–£2m. Standard.

The model existed because engineering output scaled linearly with hours, and hours scaled with bodies. Adding people was the only way to ship faster. Pyramidal leverage — many juniors per partner — was how firms made their margin.

That math does not hold any more. Code generation, retrieval-augmented research, automated test scaffolding, agent-based refactoring, and architectural pattern-matching are not incremental productivity gains for a senior engineer. They flatten the pyramid. An experienced engineer using these tools well does the bulk of the work that previously required a team underneath them.

What an AI-augmented pod actually is

It is not three engineers each using GitHub Copilot. That description is shallow and wrong.

An AI-augmented engineering pod is three senior engineers operating with:

Internal LLM tooling tuned to your codebase. Retrieval-augmented generation over your repos, runbooks, and ADRs. The senior engineer never starts from a blank context — they start with the system's history loaded.
Test scaffolding agents. Generating the boring 70% of a test suite (table-driven cases, edge enumeration, contract drift detection) so the senior writes only the interesting 30%.
Architectural review agents. Every pull request reviewed by both a senior pod member and an LLM trained on the team's design patterns. Drift gets caught before merge, not in retrospective.
Documentation agents. Runbooks, ADRs, and handoff docs written continuously, not as a deliverable at the end. The pod's institutional knowledge accumulates as code does.

The leverage here is not 30%. It is 5–10× per senior. When three seniors operate this way, a 90-day engagement produces roughly the deliverable surface area of a six-month twelve-person traditional engagement.

The math, side by side

Dimension	20-person traditional engagement	3-person AI-augmented pod
Team shape	2 senior architects, 6 mid-level, 8 junior, 2 PMs, 2 BAs	3 senior engineers. No juniors. No PMs.
Blended day rate	£600–800	£900–1,100
Monthly cost	£150,000–£280,000	£20,000–£40,000
6-month total	£900,000–£1,700,000	£120,000–£240,000
Time to first measurable delivery	4–8 weeks (kickoff + ramp)	1–2 weeks. No ramp.
Knowledge transfer	Discrete deliverable at end	Continuous, embedded throughout
Surface area shipped (6 months)	Comparable	Comparable

The cost delta is 6–8×. The output is comparable. Not "almost as good" — comparable.

Where this breaks down

It would be dishonest to claim every engagement maps to the pod model. Three places it does not:

Pure body-shop work. If the requirement is genuinely "we need 20 hands typing for 6 months on well-defined tasks," the pod is wrong. We do not do staff augmentation.
Greenfield exploratory R&D. Where the unknown is not *how* to build it but *what* to build, the multiplier collapses. AI-augmentation helps execution, not discovery. The pod works on problems with a defined target.
Highly siloed or politically fragmented orgs. The pod relies on direct engineering relationships and short feedback loops. If the path from problem to merge requires four committee approvals, the pod loses its leverage. The engagement needs an executive sponsor who can clear the path.

For everything else — modernisation, AI integration, platform reliability, delivery acceleration, cloud cost — the pod outperforms.

What 90 days with a pod actually looks like

Weeks 1–2: install. The pod connects to the codebase, deploys internal LLM tooling, maps the system. By the end of week 2, the pod knows the system at the level a Big-Five engagement reaches at week 8.
Weeks 3–6: the heavy lift. The defined goal (modernisation, integration, platform stand-up) gets shipped. Continuous delivery, daily merges, weekly demos. Your team is in every PR review — institutional knowledge transferring as the work proceeds.
Weeks 7–10: hand-off. The pod's role shifts from doing to enabling. Your team takes ownership of new patterns. Pod members move to advisory.
Weeks 11–12: exit. Runbooks, ADRs, and the AI tooling configuration handed off. The pod leaves. Your team operates independently.

This is not a hypothetical model. It is how every engagement Anystack has run for the past two years has worked.

The CTOs we lose

Two patterns of buyer rejection are honest:

The CTO who needs political cover. A 20-person Big-Five engagement is a defensible procurement choice if the project fails. A 3-person pod is not — if it fails, the CTO chose the smaller team. We understand. We are not the right partner for political risk hedging.
The CTO who wants to feel taken care of. Big consulting engagements come with a partner relationship, quarterly business reviews, account managers, and dedicated steering. We are three engineers. We do not have a CRM. If the procurement experience matters more than the engineering outcome, we are not your firm.

For everyone else: the math is straightforward.

If you want to scope a pod engagement, book a 30-minute call. If you would rather see the model applied to specific problems first, our case studies walk through outcomes at Tesco Mobile, EY, and John Lewis.

Frequently asked questions

What is an AI-augmented engineering pod?

An AI-augmented engineering pod is a small team of senior engineers (typically three) operating with internal LLM tooling — retrieval-augmented generation over client codebases, test-scaffolding agents, architectural review agents, and continuous documentation. The combination delivers comparable output to a traditional 15-20 person consulting team at a fraction of the cost.

How much does a 3-person AI-augmented pod cost compared to a traditional consulting engagement?

A traditional 20-person enterprise consulting engagement runs £150,000-£280,000 per month, totalling £900k-£1.7m over six months. A 3-person AI-augmented pod runs £20,000-£40,000 per month, totalling £120k-£240k over six months — roughly one-sixth the cost for comparable output and a faster ramp.

When does an AI-augmented pod NOT work?

Pods don't work for pure staff-augmentation (when the requirement is many hands on well-defined tasks), greenfield exploratory R&D (where the unknown is what to build, not how), or highly siloed organisations where every merge requires committee approval. The pod model relies on direct engineering relationships and short feedback loops to deliver its leverage.

How is this different from a normal engineering team using GitHub Copilot?

Copilot is incremental productivity for individual engineers, roughly a 15-30% lift on coding tasks. An AI-augmented pod is a different operating model: LLM tooling tuned to the client's codebase, test scaffolding agents, architectural review agents, and continuous documentation built into the engagement. The leverage per senior engineer is 5-10× per person, not 30%.

What does the first 90 days with an Anystack pod look like?

Weeks 1-2 the pod installs: codebase access, LLM tooling deployment, system mapping. Weeks 3-6 the defined goal ships (modernisation, integration, platform stand-up) with continuous delivery and weekly demos. Weeks 7-10 hand-off begins, with the client team taking ownership of new patterns. Weeks 11-12 exit, with runbooks, ADRs, and AI tooling configuration delivered.