Where Do Associate Hours Go Before a Sell-Side Mandate Reaches Market?

Start with the preparation boundary

Define the start and stop points before collecting times. A useful boundary might begin when the engagement team receives the initial financial package and end when the teaser, information memorandum, management presentation, or first controlled buyer-facing materials are ready for senior approval. The exact boundary should match the bank’s process. It should not silently include marketing, outreach, negotiation, or diligence performed after launch.

This matters because sell-side diligence often starts early, before marketing materials are released, and findings may inform the teaser, information memorandum, and management presentation. The work is not one uniform task. It combines analysis, evidence gathering, drafting, coordination, and review, and the mix changes with the transaction and the information available. [1]

Map work into observable task units

Use task units that an associate can start and stop without interpretation. Financial cleanup can include importing a trial balance, reconciling management accounts to financial statements, normalizing a chart of accounts, investigating unusual movements, and preparing an adjustment schedule. Document work can include building an information request, checking receipt, indexing files, following up on gaps, and linking an answer to its source.

Drafting can include preparing a company description, business model section, historical financial narrative, forecast discussion, buyer question log, or management presentation page. Review can include factual checking, tie-out, senior comments, formatting correction, and a second pass after new information arrives. Record the task output and its source-of-truth record, not just a broad label such as “deal work.”

Capture time without pretending every minute is comparable

For each task, record date, owner role, task category, start and stop or active minutes, source package version, status, and whether the work was first pass, review, or rework. Separate active work from waiting time. If an associate spends time switching between a request tracker, spreadsheet, data room, and draft, record the active task rather than assigning the whole block to whichever file was open last.

Add a short reason code for rework: changed source data, unclear request, formula or tie-out error, reviewer preference, missing context, formatting, or scope change. The purpose is diagnosis. A correction caused by newly received information is not the same as a preventable extraction error. Do not convert these records into a productivity score without agreeing how quality and complexity will be handled.

Baseline the four largest workload pools

Financial cleanup is often the first pool to measure because it creates dependencies for later drafting. Track the number and type of source files, reconciliations performed, unresolved exceptions, and time spent explaining a difference. The relevant outcome is a traceable working schedule, not merely a completed spreadsheet.

Document requests deserve their own pool. Record time to create the initial request, classify responses, identify missing support, update the tracker, and prepare follow-up questions. A request that changes several times because the scope is unclear should show that history. That record can reveal whether delay comes from the target’s records, the bank’s request design, or a review decision.

Drafting and review should not be combined. Track drafting from approved facts, review against source records, senior review, correction, and final sign-off separately. ICAEW notes that the scope and depth of financial due diligence depend on transaction materiality, user needs, and access to information and management. That is a reason to preserve context with the time record rather than compare raw hours across unrelated mandates. [1]

Create a baseline sheet that survives handoffs

A practical baseline has one row per task instance and columns for mandate stage, workstream, task description, owner role, source-of-truth link or file identifier, input version, active minutes, waiting minutes, review minutes, correction minutes, output status, and rework reason. Add a complexity note such as number of entities, periods, source files, or pages only when the team can define it consistently. Avoid invented thresholds and do not use complexity notes as a proxy for value.

At the end of each preparation cycle, summarize active time by workstream and role, review time as its own category, and rework by cause. Preserve the underlying rows. A total without the task trail cannot answer whether a future pilot reduced extraction effort while increasing review or correction effort. The baseline should be versioned so later changes to the source package do not overwrite the earlier comparison.

Choose AI trials from the measured baseline

Select a task only when its input, expected output, reviewer, and stopping condition are clear. Extraction from a defined set of financial or diligence documents may be a reasonable trial. A bounded drafting task based on an approved fact sheet may also be suitable. Keep the original source records and require the associate to verify every extracted figure, date, name, and qualification before reuse.

Do not begin with an open-ended instruction to “prepare the deal materials.” Start with one repeatable task, one source set, and one accountable human reviewer. FINRA identifies summarization and information extraction as common GenAI uses, while also emphasizing accuracy risks, documentation, testing, monitoring, and human-in-the-loop validation. [3]

Before using confidential material, confirm the firm’s approved tool, retention settings, access permissions, and contractual restrictions. If those conditions are unknown, use sanitized or synthetic material for the trial and record that limitation. NIST describes its AI Risk Management Framework as voluntary guidance for managing AI risks across design, development, use, and evaluation. It is a useful structure for assigning governance, mapping risks, measuring behavior, and managing issues, not a guarantee that a tool is suitable for a particular mandate. [2]

Measure the pilot as a complete workflow

Use the same task definition and baseline fields during the pilot. Log time for source preparation, extraction, drafting, human review, corrections, and any additional checking. Record the number and type of exceptions, unsupported statements, missing citations, formula errors, and reviewer interventions. If the pilot changes the task boundary, mark the observation as non-comparable instead of forcing it into the baseline.

Set a stopping condition before the trial begins. Examples include a repeated unsupported output, an untraceable source, a confidentiality or permission problem, a correction burden that exceeds the team’s tolerance, or a reviewer who cannot validate the output within the agreed process. Stopping is a control, not a failure. FINRA also points to prompt and output logs, model-version tracking, ongoing monitoring, and regular checks for errors or bias when firms use GenAI. [3]

Turn the record into a staffing and rollout decision

After the baseline and pilot, compare task-level distributions rather than one headline number. Ask whether extraction time changed, whether drafting time changed, whether review expanded, and whether rework shifted from one cause to another. Consider the quality of the output and the time required to establish its source trail. A task that appears faster before review may not reduce total effort once correction is included.

Use the findings to decide whether to stop, repeat with tighter inputs, or expand to another narrowly defined task. Keep human ownership for source interpretation, material judgments, approvals, and anything released outside the team. The result should be a documented operating choice with an accountable reviewer, approved data boundaries, and a next measurement period, not a promise of savings.

What the baseline should deliver

The implementation outcome is a task-level baseline of associate hours, review effort, and rework for later AI pilot comparisons. A useful handoff includes the task taxonomy, timing rules, source-version convention, reviewer roles, rework codes, pilot eligibility criteria, stopping conditions, and a short summary of observed workload by preparation stage.

Acquivo is a software platform that delivers the entire sell-side M&A process, valuation, marketing materials, buyer matching, and deal management, on one platform for business owners, M&A attorneys, and boutique investment banks. The platform is user-operated software, so a bank should still define its own permissions, source checks, reviewers, and release controls for this measurement workflow.