Recovering Revenue Per Employee at a Research and Consulting Firm

The client is a specialized research and consulting firm with ~60 employees, serving clients subscribed to research publications and corporate clients who commission bespoke advisory across financial markets, macro strategy, and regulatory themes. The firm was generating strong research and maintaining client relationships well. But a harder metric was moving in the wrong direction: revenue per employee had been declining steadily as the operations team grew faster than the research output it supported. The problem was not talent or effort — it was how analyst time was being allocated.

Analysts were spending a disproportionate share of their time finding and processing information: ingesting large volumes of policy documents, regulatory filings, macroeconomic data, and proprietary financial datasets before they could begin forming a view. Unlike larger firms that can delegate that intake work to a junior research pool or an offshore team, this firm's size meant that senior analysts were doing it themselves. The synthesis and narrative development — the work clients were actually paying for — was getting compressed.

Aegion conducted a five-day on-site assessment at the firm's US headquarters. We shadowed analysts through full working days, interviewed operations and partner-level staff, and mapped the firm's workflows end to end. With the lenses of revenue maximization and cost optimization, and with the filter of complementing the existing team and workflows, we identified several areas with a high confidence of near-term impact. We found opportunities on both sides.

On the revenue side, the constraint was straightforward: analysts weren't spending enough time on the work that differentiated the firm. Information intake consumed a disproportionate share of their day — searching for, evaluating, and processing policy papers, regulatory filings, macroeconomic publications, and proprietary data before analysis could begin. First-pass synthesis compounded this; analysts were producing multiple drafts manually before converging on a narrative. A parallel drag came from coordination work — client correspondence, meeting follow-ups, action-item tracking — handled by both analysts and operations staff rather than systematized. And years of accumulated proprietary research sat largely inaccessible, leaving analysts either recreating prior work or hunting for it through colleagues. The opportunity was to turn analysts into editors and synthesizers, not researchers and administrators.

On the cost side, the operations function had grown to support workflows that were manual by default rather than by necessity. Expense classification, invoice reconciliation, exceptions reviews, and client chargebacks were all handled by people. Proposal generation for new consulting mandates was written from scratch each time, despite following consistent structural patterns. These were not judgment-intensive activities. They were process gaps.

Every implementation decision started from the same constraint: the team had to trust the output before it went anywhere near a client. We were not building tools to automate research — we were building tools to remove the work surrounding it.

Before any of that was possible, we invested in the work that determines everything downstream: structuring the firm's own data so it was actually usable. Research archives, operational data, and client records were each restructured into clean, queryable, consistently organized sources. AI systems are only as good as the data they operate on — Aegion's edge is doing this integration work rigorously first, so that what the AI produces is trustworthy rather than merely fluent.

On the revenue side, we deployed systems to return analyst time to synthesis. An LLM pipeline ingests source material and produces structured briefs organized around the firm's research taxonomy — analysts interrogate a brief rather than read source documents in full. A drafting agent takes an analyst's notes and produces multiple narrative variants, turning analysts into editors rather than writers. A meeting capture layer handles call summaries and draft follow-ups. Years of proprietary research were made queryable through an attributed archive, retrievable in seconds.

On the cost side, we replaced manual process with structured automation. Expense classification runs against the firm's chart of accounts, surfacing only exceptions for human review. Invoice processing was similarly automated: invoices are matched against open purchase orders and reconciled, with only mismatches routed for review. Proposal generation was templatized from the firm's own historical proposals; partners work from a structured draft rather than a blank page.

Eight weeks in, the indicators are moving in the right direction across both vectors. On the analyst side, research turnaround has been cut by roughly half, and average project length has compressed by approximately a third — the firm is producing more with the same headcount. Client satisfaction scores are up 20%, with faster delivery cited as the primary driver. On the operations side, two open roles will not be backfilled; the workload is being absorbed by the systems we put in place. Early measurements suggest revenue per consultant is up 20% — a number we expect to improve with additional implementations.