AI for a Consulting Business: Research, Drafts, and Client Work

AI helps consulting teams move faster through the tasks that precede good advice: background research, framework comparisons, interview guide drafts, and slide outlines. The judgment, the client relationship, and the final recommendation stay with the consultant. Tools like fenbs keep the engagement context and the working rules your team has agreed on in a shared record, so every assistant you use starts from the current state of the project.

The gains are real but specific. A first draft of a market sizing exercise or a stakeholder communication that would take two hours takes twenty minutes. The review, the strategic layer, and the client-specific nuance still take human time.

Where AI fits into consulting work

Consulting work divides roughly into research, synthesis, communication, and judgment. AI is most useful in the first two and moderately useful in the third. Judgment remains the consultant's job.

• Background research: summarizing industries, regulatory environments, competitor profiles, and published benchmarks.
• Framework selection: comparing analytical approaches for a specific problem type and listing pros and cons.
• Interview guides: drafting questions by stakeholder type based on the engagement's objectives.
• Slide structure: generating an outline for a client presentation based on the key messages.
• Report sections: drafting background, methodology, and executive summary sections for expert review.
• Proposal templates: drafting scope-of-work language for recurring engagement types.

AI is a poor fit for tasks requiring client-specific knowledge the model doesn't have, access to proprietary data, or strategic recommendations backed by professional accountability. Those tasks stay with the consultant.

Research and background preparation

At the start of an engagement, AI can produce a working brief on the client's industry, the relevant regulatory environment, and known competitors in minutes. Give it a focused prompt: the industry, the geography, the business question, and the depth you need. The result is a starting point for your own research, not a deliverable.

Check facts against primary sources before relying on them. AI research is useful for orientation and for generating the questions to investigate further. It is not useful for precise numbers, current regulatory status, or claims you will put in front of a client. Treat it as a fast first draft of a research agenda.

For related AI approaches specific to other parts of consulting work, see AI for management consultants and the broader AI for consultants workflow guide.

Drafting client deliverables

AI is particularly useful for the structural and mechanical parts of deliverable preparation: the executive summary that must hit specific points, the options analysis table with four rows and four columns, the methodology section that explains a standard approach. Give it the key messages and the format, and it returns a working draft.

The draft will be generic. It won't know the client's preferred tone, the internal terminology that matters to their leadership, or the finding that changes everything. Your job is to add those layers. Starting from a structure is still faster than starting from a blank page.

Manage version discipline carefully. AI sessions don't retain memory of earlier drafts unless you bring it in. If a deliverable is evolving across multiple sessions, the current approved version and the open questions need to be in the record you give the assistant each time.

Keeping engagement context across sessions

Most consulting engagements run over weeks or months, with multiple team members using different AI tools at different times. When each session starts cold, the consultant re-explains the same background: what the client approved, what the working hypothesis is, and what changed last week. That re-explanation adds up.

An engagement record with the current hypothesis, the approved scope, the open questions, and the decisions already made gives each session a running start. The assistant reads the record, works from the current position, and saves what changed. The next person picks up from there rather than from wherever their memory leaves off. The how-it-works page explains how context flows between sessions.

Keep client-sensitive information out of any AI tool that isn't cleared for it. Record the engagement context in terms of objectives, decisions, and open questions, not client financials or confidential operational data.

What consultants must still own

The recommendation is yours. AI can compare three strategic options against criteria you set, but the criteria, the weighting, and the judgment about fit for this client are professional work that requires your expertise and accountability. Never present AI-generated analysis as your own judgment without reviewing it against what you actually know about the client.

Client relationships are yours. How you communicate a difficult finding, how you manage a skeptical stakeholder, and how you read the room in a working session are not tasks AI can help with in the moment. The research and the draft make you more prepared; the consultation itself is still human.

Quality control is yours. Check AI output for errors before it reaches a client. A number that is close but wrong, a regulatory claim that is out of date, or a competitor description that is inaccurate damages trust that took months to build. Build the review step into your process, not as an afterthought.

Working across a consulting team

When multiple consultants work on the same engagement, shared AI context becomes more valuable and harder to maintain. Each person's private chat history diverges from the others. Prompt conventions differ. The client's name gets used in different ways.

A shared workspace with the engagement scope, current status, open questions, and team decisions gives every team member's AI assistant a consistent starting point. When the hypothesis changes or the client narrows the scope, updating the shared record changes what every subsequent assistant sees. Roles control which team members can update which parts.

How to do this with fenbs

1. Create a free account and add a project for each active engagement.
2. Add the key context: the engagement objective, the current scope, the working hypothesis, and the decisions that are settled.
3. Write your standing working rules as Decisions and rules. For example: "Always check numbers against primary sources before including in a client document" or "Use the client's internal terminology from the approved glossary."
4. Connect your AI app from Integrations. Claude, ChatGPT, Cursor, and others connect over MCP.
5. Start each session by asking your assistant to read the engagement context. End each session by asking it to save progress: what was completed, what changed, and what the next step is.

Start with the engagement that is most active. A record covering one engagement accurately is more useful than a template covering ten poorly. Build the habit on one project, then expand.