The firm already owns most of the answer

A partner gets a new lead. It is a familiar type of client, in a familiar industry, with a problem the firm has solved before.

Then the usual process starts.

Someone searches SharePoint. Another person asks in Slack if anybody has an old deck. A manager pulls together three proposals from different folders. The partner remembers a useful framework from a project two years ago, but can’t find the final version. A consultant starts researching the industry from scratch because it feels quicker than hunting through old work.

Four days later, the team has a new proposal. It looks good. It may even win.

But a large part of that work already existed inside the firm.

For many consulting and advisory firms, around 70% of the raw material for a new proposal, diagnostic, workshop, research brief, or client deck exists in past projects. The issue isn’t that the firm lacks intellectual property. The issue is that nobody can reliably find it, assess whether it is current, and adapt it to the opportunity in front of them.

That creates a quiet but expensive operating problem. Senior people spend 20 to 40 hours on a major proposal. Consultants repeat market research that a previous engagement team completed six months earlier. Good thinking gets buried in slide decks, Word documents, email threads, workshop notes, and meeting transcripts.

The firm pays for the same insight twice. Sometimes three times.

For a consulting firm doing between $1 million and $25 million in annual revenue, we usually see the leakage from repeated delivery work land somewhere in the $80,000 to $300,000 range each year. That isn’t always a clean line item in the accounts. It shows up as lower utilisation, partner time spent editing slides, late proposals, and teams that can’t increase delivery capacity without hiring.

The answer isn’t a bigger document library. It is an operating system that makes the firm’s existing IP usable at the point of work.

Why knowledge libraries don’t solve this problem

Most firms have attempted knowledge management before.

They built a shared drive. They introduced templates. They created a folder called “Final Deliverables.” Some invested in a formal knowledge platform. A few appointed a person to maintain it alongside their main role.

Then client work got busy.

The problem is that a library asks the consultant to know what they are looking for before they search. It also asks them to decide which version is right, whether the content is safe to reuse, and how to reconcile differences between five similar documents.

That is not a search problem alone. It is a synthesis problem.

Imagine a director preparing a proposal for a mid-market manufacturer that needs a cost transformation program. The firm may have:

• A proposal from a similar manufacturing client
• A cost diagnostic framework built for a private equity operating partner
• A workshop agenda from a supply chain engagement
• A pricing model in a spreadsheet owned by one partner
• Relevant case studies hidden in completed pitch decks
• Meeting notes describing objections that tend to come up in the sales process

A folder structure cannot connect those pieces and produce a useful starting point. The director still has to do that work manually.

A well-designed AI agent can.

The goal is not to have AI blindly generate a deck. Clients can spot generic consulting language quickly, and experienced partners should not hand over judgement about scope, positioning, or commercial risk.

The goal is to give the team a high-quality first draft built from the firm’s own approved thinking. The partner then spends time making decisions, improving the argument, and tailoring the work, instead of rebuilding basic material from memory.

If you are trying to identify where this type of waste sits across your firm, See Omni for consulting firms. The audit is designed around the work that consumes real delivery and partner capacity.

Where consultants repeatedly recreate work

The issue usually starts before the engagement is sold and continues long after kickoff.

Proposal and pitch development

Major proposals are costly even when a firm has a healthy win rate.

A senior manager might pull content from old decks. A partner rewrites the executive summary. A consultant researches the prospect and competitors. Another team member finds relevant credentials and case studies. Pricing is assembled from an old spreadsheet, then revised because the old commercial structure does not quite fit.

There is nothing wrong with tailoring a proposal. The waste sits in rebuilding standard components every time.

Most firms have recurring proposal elements:

• Point-of-view language for common client problems
• Scope modules and workstream descriptions
• Team biographies and relevant credentials
• Case studies with measurable outcomes
• Workshop agendas and diagnostic approaches
• Commercial assumptions and pricing ranges
• Risks, dependencies, and client responsibilities

These should not live as disconnected fragments in individual laptops. They should be available as evidence-backed components that can be selected and adapted.

A Proposal Generation Agent in Omni ops can take a short opportunity brief, such as the prospect, sector, stated issue, team size, deal value, and required timing. It then searches approved past proposals, relevant case studies, methodology documents, and commercial models.

The first output is not a generic proposal. It is a structured draft that shows its source material. The partner can see which prior engagement informed a case study, which scope wording came from an approved service line, and where the agent could not find enough evidence.

That traceability matters. It lets the firm reuse its IP without accidentally reusing outdated claims, confidential client details, or pricing that no longer makes sense.

Research at the start of each engagement

The first two weeks of an engagement often contain the same cycle.

The team researches the sector. They identify market trends. They collect company background. They review annual reports, public filings, competitor materials, earnings calls, and trade publications. They create a market context deck or a one-page executive brief.

Some of this work must be specific to the client. Much of it does not.

A firm that has worked with retail, health, logistics, financial services, or industrial clients for years should not treat every new engagement as though it is entering that industry for the first time. The accumulated knowledge should create a faster and better starting point.

A Research Agent can run a structured research sequence at the start of an engagement. It gathers current public information, applies the firm’s preferred research framework, checks internal material for prior points of view, and produces a sourced briefing pack.

A useful output might include:

1. A one-page company and market brief
2. The client’s recent strategic priorities and financial context
3. Key competitors and market pressures
4. Relevant internal case studies and frameworks
5. A list of unanswered questions for the kickoff meeting
6. Links back to sources and internal documents

This does not replace analyst judgement. It stops analysts spending days collecting baseline facts that the firm has already collected or can gather through a repeatable process.

For ideas on where this fits with broader operating improvements, the Omni advisory approach focuses on turning isolated AI experiments into work that people use each week.

Deliverable production during delivery

The bigger cost often appears after the sale.

A team begins an operating model review, transformation program, commercial strategy project, or due diligence assignment. They need to create interview guides, diagnostic tools, workshop decks, project plans, issue trees, analysis templates, and final recommendations.

These deliverables often start with, “Does anyone have a good example?”

That question is a signal. It means the firm has useful assets but no dependable way to retrieve and adapt them.

One trades-business owner in our network described a similar issue as paying senior people to “remember where we put our own brain.” Consulting firms feel the same pain, but it can be harder to see because the work is billed by the project and buried in delivery effort.

The right standard is not that every engagement uses an identical deck. The right standard is that every consultant can start from the best available version of the firm’s existing work, then improve it for the client.

What an AI knowledge system looks like in practice

There is a practical sequence to this. It is not a matter of connecting every file the firm has ever created and hoping for good results.

1. Identify the repeatable deliverables

Start with the outputs that consume time and appear frequently. For most consulting firms, that might include:

• Proposals above a defined deal value
• Discovery briefs before new engagements
• Client research packs
• Diagnostic workshop decks
• Project charters and mobilization plans
• Case study summaries
• Standard analysis templates
• Steering committee updates

Pick one workflow first. A proposal workflow is often a strong place to start because it affects cost of sale, speed to response, and partner time.

Be clear about what “good” means. For example, a proposal draft might need a client-specific executive summary, relevant credentials, a clear scope, a delivery plan, commercial assumptions, and source references for every reused case study.

2. Prepare the source material

Not every document should enter the knowledge base.

The firm needs clear rules for approved source content, confidential material, client names, pricing access, and retention. Some prior deliverables can be used as internal references but should never be copied directly. Others may need redaction before they are available for reuse.

This is where an implementation earns its value. The issue is not loading files. The issue is setting sensible access rules and deciding what the agent is permitted to do.

The Knowledge Agent reads approved decks, documents, and meeting transcripts across the firm’s corpus. It tags and indexes material around the terms consultants actually use, such as sector, service line, client type, problem, methodology, deliverable type, outcome, and partner owner.

Then a consultant can ask direct questions:

• What have we used to assess procurement maturity in industrial clients?
• Find three relevant case studies for a CFO-led cost program.
• Which workshop agendas have worked for executive teams with limited time?
• What assumptions have we used in fixed-fee transformation proposals?
• Show me our latest point of view on post-merger integration.

The answer should include document references, not unsupported prose. The agent becomes useful because it can find and assemble context across the corpus, while the consultant retains responsibility for the client recommendation.

3. Create a review loop

AI output should go through the same professional review that applies to any work prepared by a junior team member.

For proposals, the proposal lead reviews positioning and scope. The commercial lead checks price and terms. The partner confirms client relevance. For research, someone checks sources and removes weak claims. For deliverables, the engagement manager ensures the recommendation reflects the actual situation.

The agent gets better when those edits feed a controlled improvement process. If a partner consistently changes how a scope section is worded, that is a useful signal. If a case study is outdated, it should be marked as such. If a methodology has changed, the old version should not be the default.

This is why we build agents around real workflows, not just a chat box with access to a folder.

If you want to map your first workflow and the inputs it needs, download Deploy Your First Business Agent. It is a practical worksheet for defining the trigger, source material, reviewer, output, and success measure before you start building.

You can also access the direct checklist here: Deploy Your First Business Agent.

Measure the time recovered, not just AI usage

“People used the agent” is not a useful business outcome.

A consulting firm should measure the workflow it is changing. For a proposal agent, track the time from opportunity qualification to first complete draft, senior hours spent in preparation, and the percentage of proposal content sourced from approved firm assets.

For a research agent, track days to briefing pack, analyst hours required, source quality, and the number of repeated research requests avoided.

For a knowledge agent, track the time spent searching for prior work, the number of reused approved assets, and how often teams create net-new material for a problem the firm has already addressed.

The initial goal is usually modest. If a 30-hour proposal process becomes a 14-hour process without reducing quality, that is meaningful. If four associates each recover three to five hours a week from repeated research and document hunting, the capacity adds up quickly.

You also need to be honest about where the saved time goes. In a fixed-fee environment, recovered hours protect margin. In a capacity-constrained firm, they allow more delivery without immediate hiring. In a growth phase, they give senior people more room to sell, lead clients, and develop the team.

The benefit is not that the firm produces more slides. It is that the firm stops using expensive judgement for low-value reconstruction work.

Start with one use case, then expand

A common mistake is attempting to solve every knowledge problem at once. The result is a large platform project with no working behaviour change.

Start with one high-frequency workflow, one source set, and one group of users. Prove that the agent can produce work people trust. Then expand.

A sensible order often looks like this:

1. Proposal retrieval and first-draft generation
2. Research briefs at engagement kickoff
3. Delivery templates and methodology retrieval
4. Cross-firm knowledge access through the Knowledge Agent
5. Automated updates to case studies and capability materials

The firm should also establish ownership. Someone needs to decide what counts as approved IP, which materials are current, and who can access commercially sensitive information. That does not need to be a large knowledge management department. It does need to be a real responsibility.

The broader Omni platform can support this by connecting the agent work to your existing documents, CRM data, operating processes, and review points. The technology matters, but the workflow design matters more.

Find the leakage before you buy more tools

If partners are still asking around for the latest deck, if proposal teams rebuild common sections every week, or if each engagement starts with the same research sprint, the firm has a reusable-IP problem.

It is usually bigger than the team thinks because the cost is fragmented across sales, delivery, and management time.

The first step is to identify the three deliverables that consume the most repeat effort, trace where the source material currently sits, and calculate how many senior and junior hours are being used to recreate it. That gives you a clear starting point for an agent that earns its place in the business.

Book a 60-min Omni Audit and we will work through the workflow in practical terms. In 60 minutes, you will leave with three outputs: the highest-value use case, the process and data needed to support it, and a clear view of the likely operational return. No slide deck. No vague AI roadmap.

You can also see the AI audit for consulting firms to understand how we assess proposal effort, repeated research, and knowledge management debt across the firm.

The objective is straightforward. Your best thinking should become easier to reuse after every engagement, not harder to find. When that happens, consultants spend more time solving the client’s actual problem and less time rebuilding the firm’s last answer.

Book my Omni Audit when you are ready to identify where that $80,000 to $300,000 of annual leakage is sitting in your own delivery model.