Run More Searches with Governed AI

The mandate machine: how a search firm actually grows

The principle

Strip the mystique and a retained search firm sells three things: access, discretion, and assessment judgment, delivered one mandate at a time. The economics follow. Revenue is mandates taken times the fee the firm's name commands. Capacity is the number of searches each consultant can run well at once, and the research bench sets how fast every one of them moves. Underneath both sits the real asset: a network of relationships built over a career, maintained between mandates, and lent to clients one introduction at a time. None of this scales the way a spreadsheet wants it to. A new consultant arrives with a thinner network. A bigger database is not a better lens. Growth plans stall exactly where consultant attention runs out, and the firm feels it first as searches that move slower, not as a number anyone reports.

The trap

The trap is chasing throughput with tools while leaving the operating model alone. A sourcing database here, a note-taking assistant there, and a year later the profiles arrive faster but consultants still spend evenings on market maps the firm has built before, status updates assembled by hand, and outreach admin that was never the job. The licenses were real. Capacity never moved, because nobody redesigned the workflow the tools were supposed to serve, and the constraint was never the software.

The checklist

• Name the constraint desk by desk: winning mandates, covering markets, moving candidates through process, or reporting to clients. Different problems, different fixes.
• Follow one week of consultant hours and mark which of them only that consultant could have spent. Candidate conversations and references qualify. Assembly does not.
• Count how often the firm remaps a market or rebuilds a pitch it has already produced.
• Choose workflows to improve, not tools to buy. A tool dropped on an unchanged workflow changes the invoice, not the capacity.

Where judgment beats the tool

An hours analysis shows where consultant time goes. It cannot say which hours are the product. The long lunch that keeps a twenty-year relationship warm looks like waste on a timesheet and is nothing of the kind. Deciding which hours are the fee and which are habit is a call only the partners can make.

Four outcomes that matter, and one that does not

The principle

Four outcomes justify this whole program. Expand mandate capacity without proportional hiring, so growth stops waiting on a market for proven consultants that rarely cooperates. Raise search throughput without quality slippage, so speed never arrives as narrower coverage or a staler cadence. Turn the firm's market knowledge into controlled, reusable leverage, so maps and sector insight compound instead of leaving with the researcher who built them. And adopt AI while protecting candidate confidentiality, client discretion, and assessment judgment, because those three are the fee. Notice what is not on the list: adopt AI. Adoption is a means. The moment it becomes the goal, the program starts optimizing for usage instead of for the searches.

The trap

The trap is measuring the means. Licenses issued, seats active, drafts generated: activity metrics reward the appearance of change while the operating outcomes sit unmeasured. A firm can hit every adoption target it sets and end the year with the same days to longlist, the same consultants at their ceiling, and the same client wondering why the update is late again.

The checklist

• Write each of the four outcomes as an operating sentence with a named owner, not a slide.
• Tie every initiative to exactly one outcome. An initiative that maps to none of them is a hobby with a budget.
• Baseline the outcome measures before the first pilot. An after without a before proves nothing.
• Retire activity metrics from partner reporting. Keep them in operations, where they belong.

Where judgment beats the tool

Outcomes conflict at the margin. Pushed hard enough, capacity presses on discretion, and speed presses on the care a candidate feels in the process. No dashboard arbitrates that. Where the firm sets each trade, mandate by mandate, is a leadership call, and candidates and clients experience it directly.

Who has to say yes: the six chairs in the room

The principle

Nothing durable happens in a partnership without consensus, and this program touches every chair at the table. The managing partner asks whether it grows the firm without spending its name. The COO or head of operations asks whether searches actually move faster or just generate more artifacts. Practice leaders and sector heads ask what happens to quality in markets where every candidate knows them personally. The head of technology and research operations asks whether this becomes one governed platform or a sprawl of point tools, each holding fragments of the network. The risk and data privacy lead asks what it does to candidate data, discretion obligations, and retention. The CFO asks what it costs, what it returns, and who stands behind the number. Six different questions, and the program has to hold a real answer to all of them.

The trap

The trap is the champion-led initiative that answers one chair. It moves fast on borrowed enthusiasm, then dies in a partner meeting the day the data privacy chair asks where candidate information actually goes and the room discovers nobody prepared the answer. In a consensus firm, the unanswered question does not go away. It waits.

The checklist

• Map the six chairs to named people, including the ones who hold the role without the title.
• Write each chair's question, and the evidence that would satisfy it, before the program is proposed.
• Walk the skeptics through it privately before the partner meeting ever sees it.
• Give the data privacy chair a design seat from the first day. Controls designed in read as competence; controls bolted on read as concessions.

Where judgment beats the tool

An org chart names the titles. It cannot tell you whose quiet no ends initiatives in your partnership, or whose early yes brings the undecided along. Reading the real decision structure of your own firm is judgment, and no tool has ever held it.

Research-to-Longlist: cover the market without narrowing the lens

The principle

Research-to-Longlist runs from the signed mandate to the first list a client sees: the market map, name generation, candidate research, qualification against the spec, and the longlist itself. It is where research capacity binds the whole firm, because every mandate crosses this workflow before anything else can happen, and it is where days quietly become weeks. The redesign moves assembly onto governed rails, wide before narrow, and returns researcher and consultant attention to the calls that are actually the craft: who belongs in scope, who is reachable, and who is worth the client's time.

The workflow, stage by stage

• The market map: companies and functions in scope assemble from approved sources, starting from the firm's prior work in the territory, with provenance attached and off-limits obligations checked before anyone falls in love with a name.
• Name generation: candidates surface wide from sources the firm can defend, so the lens narrows later, by judgment, instead of early, by fatigue.
• Candidate research: profiles arrive assembled with sources attached, so researchers verify and enrich instead of hunting from zero.
• Qualification: consultants judge each name against the spec, and the record keeps why each one stayed or left, which is the audit trail of the lens.
• The longlist: reaches the client in days the client can feel, with every name on it explainable.

The trap

The trap is the machine-narrowed lens. Automated research is pattern-matching, and patterns favor the already visible: the names every competitor's tool will also surface, ranked in roughly the same order. Lean on it uncritically and the firm's longlists converge on the market's default answer, which is precisely what a retained fee promises to beat. The risk is documented rather than imagined: a large field experiment on AI in professional work has been reported to show quality declining when professionals relied on it for tasks outside its capability frontier (Dell'Acqua et al., 2023; task-level results at one elite consulting firm, not a search firm), and where that frontier sits shifts with the model and the domain, which is why the consultant's read on every list is a standing control rather than a launch-phase courtesy. Faster coverage that quietly narrows the lens is not capacity. It is commoditization with better tooling.

The checklist

• Approve the source list first. Research runs on sources the firm could defend to the client, and nothing else.
• Keep off-limits and conflict checks inside the workflow, ahead of any outreach, not in anyone's memory.
• Require the wide pass before the narrow one, and record why every name leaves the list.
• Put a consultant's read on every longlist before the client sees it. Coverage is a promise the consultant signs.
• Track days to longlist and breadth of the lens together. Either one alone will flatter you.

Where judgment beats the tool

Research can assemble the market. It cannot know that the right candidate is the quiet operator a rival keeps two levels below the title, or that the name missing from every database is the one the client's board would actually follow. The lens is the consultant's, and the fee is largely paid for where it points.

Pitch-to-Mandate: win the searches worth running

The principle

Pitch-to-Mandate runs from the first conversation about a role to a signed search: qualification, pitch preparation, capability proof from prior mandates, terms, and the handoff into research. Most of that work is the firm assembling what it already knows about itself, under deadline, at partner rates. The redesign moves the assembly onto governed rails and returns partner attention to the three calls that decide everything afterward: whether to take the search, what to promise, and what the work is worth.

The workflow, stage by stage

• Qualification: what is known about the client, the role, and the firm's history in the territory arrives assembled, along with the conflicts and obligations that decide whether the firm can take the search at all.
• Pitch preparation: first drafts start from an approved record of prior mandates and sector work, not from a hunt through old decks.
• Capability proof: the track record surfaces with provenance and discretion intact, proving the work without naming what cannot be named.
• Terms: the fee conversation is informed by the firm's own record of what searches of this shape actually take to run.
• Handoff: the signed mandate becomes a structured spec, role, criteria, boundaries, and promised cadence, so the search team runs what was sold.

The trap

The trap is a pitch machine that outruns judgment. When pitching gets cheap, the temptation is to chase everything, and selectivity quietly dies. Search has a constraint consulting does not: every mandate won closes doors, because the people inside that client move off the callable market for as long as the obligation runs. A firm that wins indiscriminately is spending its own network to buy fees, and the cost never appears on any invoice.

The checklist

• Keep the take-it-or-decline gate ahead of the pitch engine. No deck starts before the decision is made.
• Surface conflicts and off-limits implications at qualification, not at signature.
• Keep capability proof in a curated record with an owner and provenance, so no pitch trades a confidence for a fee.
• Hold partner review of promise, terms, and team on every pitch. Speed is not a reason to skip the review that matters.
• Hand off every signed mandate as a structured spec of what was promised, to whom, on what cadence.

Where judgment beats the tool

Assembly can prove the firm's record. It cannot decide whether this client will honor the process, whether the role as specified can be filled at all, or what taking this search closes off elsewhere in the network. Those are partner calls, and the pitch exists to earn the right to make them.

Search-to-Client Reporting: updates a client can rely on

The principle

A retained client has bought weeks of work they cannot see, and the update is where they decide whether to believe it. Most updates are assembled late on Friday from memory and optimism. Search-to-Client Reporting rebuilds them from the live record of the search instead: pipeline status synthesized from what actually happened, outreach, responses, conversations, and declines, drafted into the update for the consultant to edit and own, on a cadence that gets kept because assembly is no longer the cost. Disclosure follows stage rules, so each candidate is visible to the client exactly as far as the conversation has earned. The update stops being a performance and becomes an instrument.

The trap

The trap is synthesis that flatters a thin pipeline. A model summarizing activity will render a stalling search in confident prose, and a client can read reassuring updates for a month before discovering the shortlist date moved. The failure is not the summary. It is a reporting chain where no claim can be challenged, because no claim traces to the record, and where declines, the most informative thing a market ever says, never reach the page.

The checklist

• Keep one live record per search, and synthesize every update from it, never from last week's update.
• Make every claim in the update traceable to the pipeline entry behind it.
• Report declines and stalls beside the progress. A spec meeting resistance is information the client paid for.
• A consultant edits and owns every update before it ships, and the promised cadence is kept especially in slow weeks.
• Set disclosure rules by stage, so no candidate is named to the client before the conversation has earned it.

Where judgment beats the tool

Synthesis can show that a search is stalling. It cannot make the call the stall usually demands: that the spec is meeting the market and the market is winning, and the client needs to hear it this week, with a recommendation. Telling a client to change the role is the moment a search firm earns its counsel. No update writes that sentence.

Candidate and client communication: discretion at the center

The principle

Every search runs two relationships at once: a client who must stay informed, and candidates who must feel respected whether they advance or not. Both run on communication, and the volume is real: approaches, scheduling, preparation, interview summaries, follow-ups, declines. AI belongs underneath that load. Approaches drafted in the consultant's own register, scheduling handled end to end, interview notes summarized into the record for the consultant to correct. The consultant keeps the relationship, the voice, and the send button. What the market experiences is a firm that is unusually responsive and unmistakably personal. What it must never experience is a machine wearing the firm's name.

The trap

The trap is outreach at machine scale. The people a search firm approaches are professionally fluent in automated flattery; they delete it daily, and they remember who sent it. One mass-personalized sequence and a relationship that took years to build reclassifies the firm as a volume shop, and the market compares notes. The network is the asset, and clumsy automation is the fastest way ever invented to spend it.

The checklist

• A consultant reads, edits, and owns every message that reaches a candidate or a client. No exceptions for the routine ones, because nobody receiving one knows it was routine.
• Draft in the consultant's own register, and cap outreach volume at what a human relationship could plausibly carry.
• Automate scheduling and logistics completely. Responsiveness is the one part of the experience a machine should own.
• Summarize interviews into the record for the consultant to correct, so the file improves without the read being delegated.
• Close every loop, especially with candidates who do not advance. This year's decline is next year's client.

Where judgment beats the tool

Drafting can produce the words. It cannot hear the hesitation that means a candidate is not truly ready to move, or judge when a week of silence is the better message, or know that this one deserves the hard feedback by phone. Reading people is the job. In this business the conversation is not overhead on the work. It is the work.

Where the risk lives: candidate data, discretion, and the read on a person

The principle

A search firm holds two confidences at once, and AI touches both. Candidates trust the firm with the most sensitive fact of their working lives: that they would consider a move, and everything they said while considering it. Clients trust the firm with departures not yet announced and strategy readable in a spec. On top of both sits the judgment the fee actually buys: the read on a person, formed in the room, and references, which are candid only because they are conversations between people who trust each other. AI raises the stakes on all of it: candidate data reaching tools nobody vetted, retention nobody decided, a summary quietly becoming a verdict. One regulatory warning and one experimental pattern anchor this chapter in evidence rather than anxiety. The U.S. EEOC and Department of Justice have warned that employers' use of AI and automated software tools in hiring, monitoring, pay, or promotion decisions can violate the Americans with Disabilities Act; that is a warning and technical assistance, not a finding about any particular tool, and nothing here is legal advice, but a search firm sits on both sides of it, screening candidates in its own process and advising clients who screen in theirs. And the field evidence on AI quality is jagged: the same experiment that reported gains on tasks inside AI's frontier has also been reported to show professional quality declining when AI was relied on for tasks outside it (Dell'Acqua et al., 2023; task-level results at one elite consulting firm, not a search firm), and where that frontier sits shifts with the model and the domain, which is why the human verdict is a standing control rather than a launch-phase courtesy. In search, the clearest outside-frontier move is letting a model make reject decisions, where an unvalidated pattern can quietly carry adverse impact at scale. None of this argues against the program. All of it argues for running the program inside boundaries that were designed, not defaulted.

The trap

The trap is treating candidate data like ordinary business records. It is not. It is career-sensitive information held on trust, much of it about people who never asked to be in anyone's file. A tenant setting does not know which mandate a note belongs to, what a spec reveals about a client's intentions, or that a referee's candor depends on never being quoted. General policy plus default settings is how a firm ends up technically compliant and actually exposed.

The checklist

• Set data boundaries at the mandate level: who is on the search defines what the tools can reach.
• Write retention for candidate data deliberately: what is kept, for how long, and what is deleted when the search closes.
• Keep the verdict human by explicit rule. Models may organize interview material; the assessment, every reference call, and every reject decision stay with the consultant.
• Check each mandate's AI posture against that client's expectations and agreements, because they differ.
• Write the disclosure answer before anyone asks, and rehearse the incident path that assumes a miss will someday happen.

Where judgment beats the tool

Controls can keep candidate data where it belongs. They cannot read a person. The half-sentence a referee declines to finish, the pause before the praise, the difference between a polished answer and a true one: the fee exists because someone in the room can hear those. Keep that someone human, and let the machines carry everything else.

The maturity path: baseline, AI-enabled, selectively AI-native

The principle

There are three honest stages, and they belong to workflows, not to firms. Baseline is where most search work lives today: maps built by hand, outreach personal, updates from memory. It is not failure; it is a ceiling. AI-enabled is the working middle: the same workflow with retrieval, drafting, and synthesis assisting inside governed boundaries, and the consultant's ownership exactly where it was. Most search workflows should live here. Selectively AI-native is the far stage: a workflow redesigned around governed AI because its volume and structure justify it. Market mapping and client reporting are plausible candidates. The read on a person, references, and the counsel around a stalling search are not, at any volume, and a firm that is clear about that list moves faster on everything else.

The trap

The familiar trap is the whole-firm transformation that burns a year and frightens the partners whose consent it needs. Search adds a quieter one: hollowing the research bench because the assembly work went away. Research years are where the next generation learns markets, patterns, and people. Redesign the work and keep the apprenticeship, or the firm trades tomorrow's judgment for this year's throughput and discovers the price much later.

The checklist

• Place each significant workflow on the path separately. The firm has no maturity level; its workflows do.
• Promote a workflow on evidence from the enabled stage, never on enthusiasm.
• Name the never-native list in writing: the read, references, and client counsel, owned by consultants at every stage.
• Redesign researcher work without deleting the learning that turns researchers into consultants.
• Revisit placements quarterly; some workflows will move in both directions.

Where judgment beats the tool

A maturity model can place a workflow on the path. It cannot decide whether the redesign is worth the disruption this year, with this bench, in this market, and it will never notice that the research seat is also the firm's school for judgment. Sequencing is strategy, and strategy is what the partners are for.

Governance that speeds the search, measured honestly

The principle

Governance here is a small set of operating patterns, each of which buys speed as well as safety. Mandate boundaries decide what AI can reach. Access mirrors the search team. Approved sources decide what research is built from, and provenance rides along, which is what makes review fast. A consultant owns every outbound word. Disclosure follows stage rules. Retention rules decide what happens to candidate data when the search closes, and reuse rules decide what the firm carries forward: the map and the lessons, never the confidences. That is the whole apparatus. Measurement keeps it honest, and six measures are enough: days from mandate signature to a client-ready longlist, searches active per consultant without quality slippage, candidate response rates to outreach, time to shortlist and time to placement by mandate type, client update cadence kept versus promised, and placement stick rate after twelve months. Baseline them before the first pilot, count program cost once, expect an adoption lag, and present ranges the partnership can believe.

The trap

The twin traps are governance as paperwork and measurement as marketing. A policy nobody opens and an approval queue people route around produce risk and bureaucracy at once. Vendor math, hours saved that nobody redeployed, benefits counted twice, payback promised for day one, produces a partnership that discounts the whole program, including the parts that were true. A firm that sells judgment is judged by its own numbers first.

The checklist

• Build controls into the path of work as defaults, and check quarterly for the ones people route around.
• Set retention, reuse, and disclosure rules at mandate close, while the boundaries are still fresh: keep the map, never the confidences.
• Baseline the six measures before the first pilot, each with one owner and one source of record.
• Count shared program cost once, price the adoption lag in, and present ranges with the assumptions attached.
• Name where freed consultant hours go: more searches, deeper coverage, or client work. Capacity without a destination quietly disappears.

Where judgment beats the tool

A pattern set can be adopted, and a measurement can prove that capacity was freed. Neither can set proportion, how much boundary a confidential succession needs versus a straightforward build-out, and neither can decide whether freed capacity becomes growth, margin, or a bench that finally breathes. Those are partner calls. The tools only make them visible.