Meridian Strategy is a four-person management consultancy in Oxford. Their clients are UK businesses with 50 to 300 employees — owner-managed firms that have grown fast and need someone outside the building to help them find operational margin. Good work, loyal clients, strong reputation. And at the start of 2026, an invisible ceiling that was costing them roughly a third of their potential revenue: 26 hours of weekly admin that no client was paying for. Three AI agents removed that ceiling. Sixty days later, the practice was delivering 61% more client projects with the same four people, 24 hours of weekly admin had been automated, and the consultancy's active pipeline had grown by more than £80,000. Here is exactly what we built.
The Constraint: 26 Hours a Week Nobody Billed For
Management consulting is a knowledge business. The value is in the analysis, the insight, and the quality of thinking in the room. But when we ran the initial time audit at Meridian Strategy, the breakdown told a different story. Across the four consultants, 26 hours a week were going into work that clients would not have valued if they could see it on an invoice.
- 11 hours on secondary research per project — benchmarking against industry peers, gathering sector data, reading competitor filings, synthesising regulatory changes. Each new engagement required the same research cycle before substantive work could begin
- 9 hours on report production — turning structured notes, meeting records, and data analysis into polished consulting deliverables. First drafts of interim reports, strategy documents, and board presentations all started from scratch every time
- 6 hours on business development — drafting capability statements, tailoring proposal documents, researching new prospects before outreach, and producing follow-up materials after initial meetings
James Mercer, the practice founder, described the situation with a clarity born from two years of tracking it: "Our billable hours are around 60% of available capacity. That is not because we cannot win the work or because the work is not there — it is because 26 hours a week disappear into getting ready to do the work and winning the next piece of it. Those hours are invisible on the P&L, but they are real."
This is the pattern we encounter consistently across professional service practices. The expertise is there. The clients want the outcome. But the operational overhead of delivering that expertise — in this case, research, reporting, and business development — compounds as the practice grows, until it becomes the binding constraint on growth. As we found in our work with a four-person Edinburgh IFA practice and a six-person Leeds solicitors' practice, the ceiling is almost always administrative rather than professional. The solution is the same: identify the three highest-volume admin processes and build an agent for each one.
The problem was not the quality of the consulting work. The problem was that every day of client work came with hours of non-billable work attached — and no obvious way to reduce it without simply doing it faster.
Agent One: The Research Agent
Every new Meridian engagement started with a research phase: understanding the sector, identifying peer benchmarks, pulling regulatory context, reviewing publicly available information about the client's market position. This research took between 8 and 14 hours per project, depending on how niche the sector was. It was necessary. It was not high-margin work.
The research agent operates from a rolling knowledge base of sector intelligence — reports, company filings, industry body publications, and analyst summaries — updated weekly via a scheduled ingestion process. When a new engagement is confirmed, the lead consultant completes a brief structured intake form: client sector, company size, primary challenge, competitors to benchmark, and the regulatory framework relevant to the engagement. The agent takes this brief and returns a structured 12-page research briefing within 90 minutes.
The briefing covers: sector overview and relevant trends, three to five peer comparisons with operational benchmarks, applicable regulatory requirements, a summary of publicly available financial or operational information about the client, and a list of open questions the briefing cannot answer from available data. That last section matters — the agent is explicit about the boundaries of what it can and cannot determine, which prevents the overconfidence problem that undermines less well-constrained AI research tools.
We built the knowledge base using the Meridian team's own prior research archives — 140 past project research files, reorganised, cleaned, and loaded as structured documents into a vector store. As we covered in the piece on AI agent memory architecture, a well-constructed retrieval layer dramatically improves output quality by ensuring the agent draws from authoritative, vetted sources rather than attempting to recall information from its training data. Every new engagement adds to the knowledge base, compounding its value over time.
Time saving: 11 hours per project reduced to under 90 minutes. Across the practice's project cadence, approximately 9 hours saved weekly.
Agent Two: The Report Writing Agent
Consulting deliverables are not generic documents. They carry the firm's voice, the client's context, and the specific analytical conclusions the engagement has produced. They cannot be generated from a blank prompt and published without review. But they absolutely can be generated from a detailed structured input — and that distinction is what made the report writing agent viable at Meridian.
The workflow is this: after any significant piece of client work — a discovery workshop, a data analysis sprint, a series of stakeholder interviews — the lead consultant spends 20 to 30 minutes completing a structured notes template. The template captures the key findings, the conclusions and recommendations, the supporting evidence, any data points to include, and the intended audience and decision context. The agent takes this template and produces a first draft of the full deliverable in the practice's standard format within 20 minutes.
The first draft is not the final document. Meridian's consultants spend 45 to 60 minutes on review, refinement, and the additions that require genuine professional judgement. What has changed is that they no longer start from a blank page. They start from a well-structured draft that captures the right architecture, the right level of formality, and the essential content — and they shape it from there. The difference in cognitive load and calendar time is significant.
The knowledge base for the report writing agent contains Meridian's own deliverable library: 85 past client documents, redacted and structured as style examples. The agent applies the firm's voice consistently — the way it frames recommendations, the structure it uses for board-ready summaries, the level of rigour applied to financial exhibits — regardless of which consultant is leading the engagement. This is the same principle we described when explaining how MCP integrations extend what agents can access: the value comes from connecting the agent's capability to your specific institutional knowledge, not from using it as a generic writing tool.
Time saving: 9 hours per project reduced to under 1.5 hours of consultant review. Across the practice's output volume, approximately 8 hours saved weekly.
Agent Three: The Business Development Agent
The third agent addresses the most neglected part of any knowledge firm's operations: business development. At Meridian, BD was happening — but it was happening between everything else. New enquiries were answered when there was time. Proposal documents were written in evenings. Prospect research before a first meeting was often cursory because the meeting was tomorrow and a client report was due today.
The BD agent runs three separate functions. First, inbound enquiry response: when a new prospect contacts the practice via the website or email, the agent sends an initial reply within 15 minutes, asking the structured qualifying questions that determine whether the enquiry is likely to be a good fit — sector, company size, the nature of the challenge, the rough timeline, and whether budget has been considered. This information is what the Meridian team needed before deciding how to proceed. Now it arrives before the first human conversation rather than consuming it.
Second, prospect briefings: before any new business meeting, the agent prepares a one-page brief — company overview, recent news, financial position from public records, known sector challenges, and a summary of any prior contact with the practice. The brief takes 12 to 15 minutes to compile. It makes every first meeting more focused and more credible, which directly improves conversion from initial conversation to formal engagement.
Third, proposal drafting: when a qualified prospect requests a proposal, the agent produces a tailored capability statement and engagement outline using Meridian's proposal library and the intake information gathered earlier. The draft captures the right scope, the relevant experience, and the appropriate fee structure — ready for the lead consultant to review and finalise in 30 to 40 minutes rather than build from scratch over two to three hours. This connects directly to the approach we described in the post on building an AI proposal writer for service businesses: the agent does not win the business, it removes the hours of production work that were preventing the humans from winning it.
Time saving: 6 hours per week of BD admin reduced to under 1 hour of review and finalisation. Approximately 5 hours recovered weekly.
The Results After 60 Days
Sixty days after deployment, the numbers at Meridian Strategy looked like this:
- 61% more client projects handled per quarter — from an average of 7.5 concurrent project tracks to 12.1, same four-person team
- 24 hours of weekly admin automated across the three agents — research (9h), report production (8h), BD admin (5h), residual scheduling and coordination (2h)
- Billable capacity up from 60% to 83% — the 26-hour admin burden reduced to under 4 hours, freeing 22 hours a week for client-facing and billable work
- Pipeline value increased by £82,000 — faster, better-quality proposal responses and improved prospect research lifted enquiry-to-engagement conversion by 22%
- Deliverable cycle reduced by 68% — documents that previously took 3 to 5 days from structured notes to client-ready now take 1 to 2 days
- Zero additional hires required to absorb the additional project volume — all capacity came from the admin hours recovered
James Mercer's assessment at the 60-day mark: "We are doing more work and doing it better. The research is more thorough because the agent covers ground we would have had to shortcut. The deliverables are more consistent because we are all working from the same structural foundation. And we are winning more proposals because we are responding faster with more specificity. I did not expect all three of those things. I expected to save some time. I got a better business."
The agents did not make Meridian better consultants. They made Meridian a better consulting business — one that could apply its expertise to more clients, produce deliverables faster, and win more work without the BD effort consuming the team.
What This Cost and What It Pays Back
The total build cost for the three-agent AI operating system at Meridian Strategy was £3,900 — a single engagement covering knowledge base construction, agent configuration, email and scheduling integrations, the BD automation layer, testing, and a 30-day post-deployment refinement period.
Ongoing running costs are approximately £90 per month: cloud compute, language model API usage, vector store hosting, and the scheduling integration. No per-seat licensing. No annual contract. The system runs continuously at the same fixed cost whether Meridian is running 7 concurrent projects or 20.
The return on that investment arrived from three directions:
- Increased project throughput. 61% more concurrent project tracks at the same average engagement value represents a significant revenue increase with no additional headcount. For Meridian, with typical project values between £18,000 and £35,000, the additional capacity translates directly into material annual revenue growth.
- Pipeline expansion. The £82,000 pipeline increase from improved BD response quality and speed represents revenue that was previously being left on the table — enquiries that moved on because Meridian was slow to respond substantively, or proposals that did not demonstrate enough specificity to close. The BD agent closed both gaps.
- Higher-quality output at scale. The consistency improvement in deliverable quality — from both the research depth and the report writing agent — has effects beyond efficiency. More thorough analysis makes for better recommendations. Better recommendations produce stronger client outcomes. Stronger client outcomes generate referrals and renewals. The agents improved the quality of the work and the capacity to do it simultaneously.
The £3,900 build cost was recovered within the first completed engagement under the new system. The pattern is consistent across every professional service practice we have built for — recruitment agencies, accountancy firms, HR consultancies, solicitors' practices. The constraint is almost always capacity, not demand. Remove the capacity constraint and the revenue follows.
If you run a management consultancy, a strategy practice, or any knowledge business where your team is spending meaningful time on the research, reporting, and BD work that surrounds your expert work — get in touch. We scope and build AI operating systems for UK service businesses in a single engagement, and we can tell you within a short call whether the numbers are likely to stack up for your practice.