AI for Consulting Firms: Win Back the Hours You Never Bill

Key Takeaways

• The leak is structural. A London Business School study found around 15% of chargeable consulting work never gets billed, a meaningful and recurring leak.
• Non-billable work is the bigger pool. Proposals, pre-meeting research, status reports, and internal admin eat hours no client ever sees on an invoice.
• An AI employee takes the wrapper, not the advice. Research packs, first-draft proposals, recaps, and status updates get drafted; the consultant keeps every recommendation.
• It works across the engagement lifecycle. From pre-sale research to kickoff to weekly client updates to the closeout summary, the same AI employee carries the context.
• Review-first fits client work. Nothing reaches a client without a consultant approving it, which is the only acceptable bar in a relationship business.
• The win is margin, not headcount. Same team, more billable share per week, faster turnaround between meetings.

Consulting has a math problem nobody puts in the deck. A London Business School study found that around 15% of chargeable client work never gets billed. And that is just the chargeable side. The hours spent on proposals, pre-meeting research, and status reports were never going on an invoice in the first place.

A firm cannot fix that by telling people to work more. The hours are already worked. This post is about handing the non-billable wrapper of consulting to an AI employee so the billable core gets the week back.

Where do a consultant's hours actually leak?

Walk an engagement from first call to closeout and the unbilled work shows up at every stage:

• researching the prospect before the pitch: company, market, competitors, recent moves
• drafting the proposal and rebuilding the same sections from the last one
• preparing for every client meeting, then writing up every client meeting
• the weekly status email each workstream owes each client
• chasing inputs the client promised two weeks ago
• the closeout summary and the case study nobody has time to write

None of this is the advice. All of it surrounds the advice, and most of it has the same shape every engagement: gather, structure, draft, send. That shape is exactly what an AI employee does well.

There is also a second-order cost that never shows up in time tracking. The consultant who spends Tuesday evening assembling a research pack walks into Wednesday's meeting tired, and the partner who writes status emails on Friday afternoon is not selling the next engagement. The leak is not just hours, it is the quality of the hours that remain.

What does an AI employee do across an engagement?

Viktor is an AI employee that lives in Slack and Microsoft Teams and connects to 3,200+ tools, which in a consulting stack means the CRM in HubSpot, email in Gmail, the document store in Notion, the project tracker in Linear, and the meeting notes in Granola. Across the lifecycle:

The pattern in the right column is deliberate. Anything a client is paying for stays human. The full version of that split for adjacent businesses is in AI for agencies, and the recurring-report mechanics are the same ones we covered in Replace weekly reporting with AI.

What does this look like before a client meeting?

The pre-meeting research pack is the cleanest single example, because every consultant builds one and almost none of that time gets billed:

@Viktor I have a meeting with the Meridian Logistics CFO on Thursday.
Pull our full history with them: HubSpot notes, email threads from the
last 90 days, the proposal we sent in March, and open items from the
project tracker. Add a one-page brief on their recent news and hiring.
Post it all in #meridian-prep by Wednesday 3pm for my review.

That is forty minutes of gathering turned into one message, and the consultant walks in with better context than the version they would have assembled at 11pm the night before. Multiply by every meeting on the calendar and the recovered hours stop being a rounding error.

How does review-first work with client deliverables?

In a relationship business there is exactly one acceptable answer: nothing reaches a client without a consultant approving it. Viktor works review-first by default, which maps cleanly onto how firms already work:

• status emails are drafted in the channel, the engagement lead approves and sends
• proposal sections arrive as drafts in the document, the partner edits and owns
• meeting recaps are posted internally first, then shared once checked
• input-chasing nudges to clients are suggested, not sent

The approval step is not overhead. It is the same junior-to-senior review loop firms have always run, with the first draft arriving in minutes instead of days. The argument for keeping that default is in Don't let your AI agent act without asking.

What should a consulting firm never delegate?

The line is the same one clients are paying to have drawn well:

• The recommendation. The analysis can be assembled by an AI employee; the advice is the product.
• The relationship. No AI sends difficult news, negotiates scope, or manages a stakeholder.
• The commercial terms. Proposal text can be drafted; the deal itself is a partner call.
• Anything under privilege or NDA-sensitive judgment. The AI employee handles documents within your access rules; deciding what a client may see stays human.

Firms that blur this line lose the thing they sell. Firms that hold it get the opposite effect: more partner hours on judgment, because fewer partner hours go to formatting.

How should a firm roll this out?

Start with one engagement team, not the whole firm. A firm-wide rollout invites a firm-wide debate, while a single team produces a result you can measure in a month. Give that team three tasks in week one:

1. pre-meeting research packs for every external meeting
2. the weekly status draft for each active client
3. meeting recaps with action items, posted to the engagement channel

All three are internal-facing or review-gated, all three are easy to verify, and all three free hours immediately. The team will know within days whether the drafts are good, because they used to write them by hand. After two weeks, compare the team's billable share and turnaround times against the previous month, ask the engagement lead what they stopped doing manually, and decide what to add next. If you want the general framework for picking what comes after, What is an AI coworker? covers how the category fits into a services business.

Frequently Asked Questions

Can an AI employee really save billable time at a consulting firm?

Yes, by attacking both leaks: the chargeable work that goes unbilled because it feels too small to record, and the non-billable wrapper of research, proposals, and status reports. The advice stays human; the gathering and drafting around it gets delegated.

Is client data safe with an AI employee?

Viktor works inside your existing tool permissions, drafts rather than sends by default, and keeps engagement context within your workspace. The firm decides per task what runs review-first, and client-facing output should always require approval.

What is the best first task for a consulting team?

Pre-meeting research packs. Every consultant needs them, the inputs already live in your CRM and email, the output is easy to verify, and the time saved shows up in the same week.

Does this replace junior consultants?

It replaces the worst part of their job. The gathering and first-formatting work that juniors burn evenings on gets drafted for them, and they move up to the review and analysis work they were actually hired to learn.

How does an AI employee handle multiple clients without mixing context?

Each engagement runs in its own channel with its own context: CRM records, email threads, documents, and trackers for that client. Briefs are scoped per channel, so the Meridian recap never pulls from another client's files.

Where should a consulting team start?

Pick one recurring, non-billable task that already eats real hours, the weekly status email or the standing prospect-research pack, and hand that over first. Run it review-first for two weeks so the team builds trust in the drafts, then expand to the next recurring job. The value compounds because the wrapper work, proposals, status reports, CRM updates, and pre-meeting research, is recurring, so every hour it takes off a consultant's plate comes back the following week too.