The hidden project cost starts after the interview

Most consulting firms know the pattern.

A partner, manager, or subject matter expert has a 45-minute conversation with a client leader, industry operator, or technical specialist. The discussion is useful. It contains the detail the project team needed. There are objections, examples, operating facts, and language that will shape the final recommendation.

Then the recording lands with a junior consultant.

They download the file, run it through a transcription tool, clean up speaker labels, replay unclear sections, and turn a 45-minute discussion into three to six pages of notes. After that, they write a summary. They pull out themes. They compare the interview against the hypothesis tree. They add quotes to a working deck. Someone senior reviews it because the important nuance is rarely captured perfectly on the first pass.

That cycle often consumes 5 to 10 hours of junior consultant time for a single expert interview. Across a discovery phase with six to 12 interviews, the total can quickly become 30 to 80 hours.

The issue isn’t that junior consultants shouldn’t learn how to listen, assess evidence, and form judgement. They should. The issue is that your most expensive developing talent is spending a meaningful part of every project doing repetitive production work. Meanwhile, the knowledge created in the interview usually becomes trapped in a project folder, a PowerPoint appendix, or personal notes.

For a consulting firm doing $1 million to $25 million in annual revenue, that leakage doesn’t appear as one clean line item. It shows up as project margin pressure, late working sessions, inconsistent outputs, and partners needing to re-read source material because they don’t fully trust the synthesis. Across the firm, we usually see the avoidable annual drag land somewhere in the $80K to $300K range.

Automating expert interview transcription and synthesis won’t remove the need for consultant judgement. It gives that judgement better inputs, faster.

What junior consultants are actually doing today

“Transcription” sounds like one discrete task. In practice, it is a chain of manual steps, each with its own failure point.

First, the team has to collect recordings. Some sit in Zoom. Others are in Teams, a phone recording app, or a specialist interview platform. Files are often named inconsistently. Consent requirements differ by client and region. The team might have to check who attended, which version is final, and whether the call contains sensitive commercial information.

Then comes the transcript. Generic tools can produce text, but consulting interviews are difficult material. Speakers interrupt each other. Acronyms appear without explanation. Industry language is specific. A speaker may say “gross margin” when they mean contribution margin. A client executive may use an internal programme name that matters to the recommendation but means nothing outside their business.

The raw transcript still needs work.

A junior consultant typically:

• checks terminology, names, and speaker attribution
• identifies the questions asked and where the answer sits in the recording
• extracts evidence against each project hypothesis
• separates facts from the interviewee’s opinions
• captures direct quotes that can safely be used in the working team
• flags contradictions with other interviews
• writes a concise interview summary
• updates the findings tracker and project deck
• prepares follow-up questions for the next call

That is real analytical work mixed with clerical work. The distinction matters. If the task is simply summarised, the team risks losing the evidence trail. If it is copied too literally, senior stakeholders won’t read it. The useful output is a structured synthesis that preserves what was said, shows why it matters, and gives the project team a clear next action.

The manual process also creates variation. One consultant writes a crisp one-page summary. Another creates seven pages of notes. One captures quotes precisely. Another paraphrases too aggressively. One team codes interviews against themes. Another works from memory in a workshop. You can still deliver good client work this way, but you pay for consistency every time.

This is closely tied to the knowledge management debt inside many firms. Each engagement produces insight, yet almost none of it becomes accessible when the next engagement begins. Our Omni ops work is designed around fixing repeatable operating processes like this, not simply putting a chat interface over a pile of files.

What an expert interview agent does end to end

An AI agent for expert interviews should be designed as a controlled workflow, not a generic summarisation prompt.

The workflow begins before the call. The agent receives the project context, workstream, interviewee role, research questions, and interview guide. It knows the hypotheses the team is testing. It also knows the output format your firm uses, such as a one-page interview brief, evidence table, quote bank, and follow-up question list.

After the interview, the workflow performs a defined sequence.

1. Capture and organise the source material

The agent picks up the approved recording and its associated meeting details. It applies a standard naming convention and links the interview to the correct client, engagement, phase, and workstream.

This isn’t glamorous, but it is essential. A strong process starts by preventing the confusion that appears when files are distributed across SharePoint, Google Drive, Teams, email, and individual laptops.

It can also record consent status and access rules. For advisory firms working with confidential commercial material, this is part of the design, not an afterthought. The right implementation respects the client’s security requirements, retention policy, and permissions structure.

2. Create a transcript that can be checked

The agent produces a timestamped transcript with speaker labels. It identifies low-confidence passages and terms that need review. Rather than pretending every word is accurate, it gives the project team a short exception list.

A junior consultant might spend 20 minutes checking ten uncertain passages rather than two hours replaying an entire recording. They remain accountable for quality, but their effort is focused where it matters.

The transcript can preserve timestamps behind each finding. If a manager asks, “Where exactly did the COO say that?”, the team can go back to the source in seconds.

3. Extract evidence against the project questions

This is where a well-built agent moves beyond a basic transcription product.

It reads the transcript in the context of your interview guide and project hypotheses. It identifies statements relevant to each question, classifies them as evidence, opinion, example, risk, or open issue, and connects them to the appropriate workstream.

For example, in an operating model review, a transcript might contain comments about decision rights, reporting cadence, process bottlenecks, technology constraints, and leadership behaviour. The agent doesn’t just produce a broad summary. It groups the comments into the themes the project is already trying to test.

It should also preserve the distinction between:

• what the interviewee directly stated
• what the agent inferred as a possible implication
• what requires validation through data or another interview

That distinction protects the quality of your analysis. Consulting teams get into trouble when an interesting comment becomes an assumed fact in the deck.

4. Produce a usable interview brief

The immediate output should be something a manager can read in five minutes.

A useful brief often includes the interview purpose, participant role, top findings, supporting evidence, direct quotes, points of disagreement, implications for the workstream, and recommended follow-ups. The format should fit your existing project routines.

The agent can generate a longer evidence pack too, but it shouldn’t force senior people to read it. The best workflow creates a short decision document and retains the detailed source material behind it.

5. Update the shared knowledge base

This is where the return compounds.

The interview shouldn’t disappear once the phase ends. The agent stores the transcript, brief, tagged themes, and approved findings in the firm’s knowledge environment. A future project team can then ask questions across previous interviews, decks, and research documents, subject to the access rules you define.

That is a core role of the Knowledge Agent (Omni ops). It reads the decks, documents, and meeting transcripts your firm produces and helps people find relevant prior work without relying on who happens to remember it.

You can learn more about Omni for consulting firms and how we identify workflows that are both valuable and safe to automate.

The human role does not disappear

Some partners hesitate here for a reasonable reason. Expert interviews are not administrative transactions. A good consultant hears what wasn’t said, notices a shift in confidence, and understands the political context behind a careful answer.

AI won’t replace that.

What it can do is take the first pass through the material with discipline. It can apply the same template every time, capture evidence with timestamps, compare one interview with another, and create a structured draft before the project team meets to discuss it.

The consultant’s role shifts toward work that actually benefits from experience:

• testing whether the interviewee’s claim is credible
• recognising strategic or organisational implications
• deciding what evidence belongs in a client recommendation
• challenging assumptions
• determining the next interview or data request
• framing the story for a client audience

One trades-business owner in our network described a similar shift as “getting my team out of note production and back into the work clients hired us to do.” That is the objective. Not unattended output. Better leverage.

The workflow should include review gates. A junior consultant can approve the transcript corrections. A manager can approve key findings before they enter a client-facing deck. Sensitive content can be restricted to the core project team. The agent should make the process more auditable, not less.

Where the economics become meaningful

It is easy to focus only on time saved per interview. Say an agent removes four to seven hours of handling time from each interview. On a project with eight interviews, that may recover 32 to 56 hours.

That is useful, but it is only the first layer.

The larger gain comes when the firm standardises the process across its projects. Interview briefs become consistent. Managers spend less time asking for rework. Project teams find past evidence faster. New consultants get a more reliable model of what good synthesis looks like. Partners can review findings at a higher level because the evidence is already organised.

The impact also connects to the other repeatable work inside an advisory firm.

The Research Agent (Omni ops) can create sourced industry and company briefs at the start of an engagement, reducing the repeated secondary research that happens across client teams. The Proposal Generation Agent (Omni ops) can pull from prior proposals, case studies, and pricing material to create a tailored first draft, which helps when senior people are spending 20 to 40 hours on each major proposal.

These workflows should not be built as isolated experiments. The interview agent should connect research, delivery, and reusable firm knowledge. A research brief can inform the interview guide. Interview findings can update the project evidence base. Approved project materials can strengthen the next proposal.

If you want to map that opportunity against your own engagement model, Book a 60-min Omni Audit. We will spend 60 minutes looking at the actual work, then leave you with three things: a prioritised workflow view, an estimate of value and effort, and a practical next-step plan. No deck, no theatre.

How to choose the right first interview workflow

Don’t start by trying to automate every conversation in the firm. Start with one interview type that has enough volume, a repeatable structure, and a clear owner.

For many firms, the best candidate is discovery or diagnostic interviews in a common service line. The guide is broadly similar. The project team produces standard outputs. The work is frequent enough to show value within a few engagements.

Before building, answer these questions:

1. Which interviews recur often enough to justify a workflow?
2. What exact output does the manager need after each call?
3. Which findings must be linked to source timestamps?
4. What data, client confidentiality, and retention rules apply?
5. Who checks the output before it becomes part of the project record?
6. Where should approved knowledge live so another team can find it?
7. How will you measure success after 30, 60, and 90 days?

Avoid a vague brief like “summarise all calls.” It produces vague output.

A better instruction is: “For each operational diagnostic interview, create a one-page summary using our template. Extract evidence against the seven diagnostic questions. Include no more than five direct quotes, each timestamped. Flag contradictions with prior interviews. Separate interviewee opinion from verified facts. Recommend up to three follow-up questions.”

That is something a team can review and improve.

If you need a practical way to scope the first agent, our Deploy Your First Business Agent guide is built as a worksheet. It helps you define the trigger, inputs, decisions, review points, and measures before you build. You can access the direct workbook here: download the first-agent worksheet.

Build for adoption, not a clever demo

A working demo can transcribe a call and create a summary in minutes. Adoption requires more.

Your team needs to know where recordings go, when the agent runs, who receives the brief, and what they are expected to review. The output needs to fit into existing project management habits. It should not create another dashboard that nobody opens after week two.

Start with a small operating standard:

• a defined set of interview templates
• one approved storage location
• a standard brief format
• a review owner for each workstream
• clear rules for sensitive interviews
• a feedback mechanism for improving prompts and classifications

Then measure the right things. Track handling time per interview, time from call completion to usable brief, manager rework, and the percentage of findings with source evidence. You can also track reuse. If a team can locate prior interview evidence during proposal or project setup, the firm is beginning to reduce knowledge management debt.

The wider Omni platform is built to support this kind of operating model. The goal is not a collection of disconnected AI tools. It is an agent workflow that does defined work, connects to the right sources, and gives people visibility over what happened.

You can also browse our practical AI guides when you are comparing use cases across proposal development, research, operations, and knowledge retrieval.

Make interview synthesis a firm asset

The manual interview process looks small when viewed one call at a time. Across dozens of engagements, it becomes a predictable drain on delivery capacity and a reason your best insights remain fragmented.

A properly designed agent can give each project team faster interview briefs, clearer evidence, better follow-up questions, and a searchable record of what the firm has learned. Your junior consultants still do the thinking. They just stop spending so many hours producing first-draft notes from scratch.

The next step is to look at your own project phases, interview volume, systems, and confidentiality requirements. See Omni for consulting firms to understand the audit process, or Book my Omni Audit if you are ready to identify the first workflow worth building.