Your clients have already bought the AI

The signal matters because it changes the commercial position of a consulting firm.

A report that OpenAI enterprise revenue has exceeded consumer revenue for the first time, reaching $40 billion in annual recurring revenue two quarters ahead of expectations, is not just an AI industry headline. It is evidence that corporate buyers are no longer waiting for their advisers to introduce them to generative AI.

They have enterprise accounts. Their teams are testing ChatGPT and other tools. They have security, procurement, legal, data governance, and adoption questions already sitting on their desks. In many cases, they are using AI informally before leadership has decided where it belongs in the operating model.

That creates a clear choice for a consulting or advisory firm.

You can keep selling strategy presentations about what AI might do. That work will become harder to defend as clients become more familiar with the tools.

Or you can become the implementation partner who helps clients apply AI to the actual work that determines cost, speed, quality, and growth. That means identifying repetitive workflows, creating governed agents, connecting them to usable knowledge, and measuring whether they improve an economic outcome.

For a $1 million to $25 million consulting business, this isn’t only a client opportunity. It is a firm-building issue. Your own team is probably carrying a sizeable amount of manual work that should already be structured, searchable, and reusable.

We usually see an annual leakage range of $80K to $300K in firms of this size. It rarely appears as one large expense line. It shows up in partner time spent rewriting proposals, managers rebuilding research packs, and teams recreating insights they delivered to another client six months ago.

The same systems you need to advise clients on are the systems you should use inside your own firm.

The old model was selling expertise by the hour

Consulting has always had an understandable tension. Clients buy your experience, but your operating model often requires highly paid people to perform repeatable tasks before they can apply that experience.

A major proposal is a good example.

A new opportunity arrives through a referral, a past client, or a formal tender. A partner has an initial conversation. Someone asks for the scope, the approach, the team, comparable work, a timeline, commercial terms, and a point of view on the client’s issue.

Then the internal scramble begins.

A senior consultant searches old folders for the closest proposal. A business development manager asks three people for case studies. A director rewrites the methodology section. Someone finds an old pricing model and tries to adapt it. The partner reviews the deck at 10pm because the deadline is tomorrow.

Twenty to 40 hours on a meaningful proposal is common. The firm may still win at an acceptable rate, which makes the process easy to tolerate. But cost of sale keeps rising. Senior billable capacity gets absorbed before the engagement is even signed. The knowledge created in that proposal gets saved in another folder, usually with no clear standard for reuse.

The same pattern exists in delivery.

A client asks for a market assessment, operating model review, technology strategy, transformation roadmap, or commercial due diligence. The firm starts with secondary research. Analysts search company sites, annual reports, competitor material, trade publications, and past client work. They build a summary, then a manager rebuilds that summary into a briefing document.

Some research must be fresh. That’s part of good advisory work. But much of the process is repeated. Your firm has likely paid multiple times to find, read, classify, and summarise information it already collected for adjacent projects.

Then there is the knowledge management problem. Every engagement produces slides, documents, models, notes, interview transcripts, deliverables, and client feedback. The useful IP lives across SharePoint, Google Drive, Teams, email, personal folders, and old project workspaces. It is technically stored but not operationally available.

That is not a lack of intelligence. It is a retrieval and workflow problem.

Clients buying enterprise AI are beginning to expect their advisers to solve problems like this. If your own delivery model cannot demonstrate it, your positioning will feel theoretical.

The implementation partner position is more valuable

The client does not need another consultant to say that AI can improve productivity. Most executive teams already accept that premise.

They need help answering harder questions:

• Which workflows should be automated first?
• Where does human review stay mandatory?
• What data can an agent access?
• How should the team verify sources and outputs?
• Which measures prove that the work is creating value?
• How do we avoid rolling out disconnected tools that no one owns?

This is where a consulting firm can remain central. The enterprise AI platform is not the consulting offer. It is infrastructure. The value sits in workflow design, domain context, change adoption, process controls, and economic measurement.

A strategy firm might create an agent-assisted market intelligence process for a client. An operations consultancy might build a structured intake and issue-triage agent. A financial advisory practice might develop a review workflow that prepares evidence, identifies exceptions, and routes decisions to the right person.

The client still needs judgement. It needs someone who understands its operating model and can establish the rules around the tool. That is implementation work, and it is much closer to measurable client value than a generic AI presentation.

There is one catch. You need a credible internal example.

If you want to sell clients an AI-enabled research operating model, your own research should not begin with an analyst staring at a blank browser tab. If you want to advise on knowledge reuse, your own case studies and delivery lessons should be retrievable in minutes.

You can see how this applies to the sector through the AI audit for consulting firms. The starting point isn’t a broad software discussion. It is a close look at where work gets stuck, repeated, or handed between people without a clear system.

Start with work your firm already understands

The right first agents are not speculative projects. They should target work your team performs every week, where inputs are identifiable, the process has repeatable steps, and an experienced person can review the output.

At Omni, we commonly start with three patterns for consulting firms.

Proposal Generation Agent

The Proposal Generation Agent pulls from past proposals, case studies, scope documents, methodology pages, pricing guidance, team bios, and approved commercial language. It creates a tailored first draft for a live opportunity.

This doesn’t mean giving an AI system permission to send proposals on its own. A proposal is a commercial commitment and needs partner judgement. The agent handles the assembly work so the partner can focus on the difficult parts: the client problem, point of view, scope boundaries, deal risk, and price.

A practical flow looks like this:

1. The proposal lead completes a structured opportunity brief with the client name, problem, industry, buyer role, expected outcomes, timing, and known competitors.
2. The agent searches approved internal materials for relevant case studies, capability statements, and comparable approaches.
3. It produces a proposal outline, draft executive summary, recommended workstreams, indicative team shape, and a list of assumptions needing review.
4. It flags where the firm has no suitable evidence rather than inventing a case study.
5. The commercial lead reviews the draft, adjusts pricing and commitments, then routes it through the normal approval process.
6. The final proposal and the outcome are captured back into the knowledge base.

The gain is not only speed. It is consistency. Your firm stops relying on whoever happens to remember the best previous example. Newer staff can prepare a strong first version without pretending to have the partner’s experience.

Research Agent

The Research Agent runs structured industry and company research at the start of each engagement. It provides sources, summaries, and a one-page brief that a consultant can validate before the kickoff meeting.

This is useful because research work is often necessary but poorly designed. The assignment may be, “Get up to speed on this company and market.” That instruction creates inconsistent output. One analyst provides a few screenshots and links. Another produces 40 pages. Neither necessarily answers the questions the project team needs.

A better workflow starts with a defined research template. For example:

• Company profile and operating footprint
• Revenue model and customer segments
• Recent strategic moves
• Relevant financial and operational indicators
• Competitor and market context
• Known technology and AI activity
• Potential interview questions
• Source list and confidence notes

The Research Agent follows that structure. It identifies sources, separates fact from inference, and produces a briefing pack with links back to evidence. The engagement manager decides what matters and what requires further validation.

That last point matters. AI should reduce the time spent collecting and organising information. It should not remove professional accountability. In consulting, an unsupported assertion can damage trust quickly. The workflow needs source visibility and a named reviewer.

Knowledge Agent

The Knowledge Agent reads the decks, documents, meeting transcripts, approved frameworks, proposals, and project summaries your firm produces. It answers questions across that corpus with references to the underlying material.

This is how you reduce knowledge management debt.

Imagine a partner asks, “What have we done in the past two years for mid-market manufacturers on operating model redesign, and which outcomes can we substantiate?”

Without a usable system, someone sends messages around the business. People search project folders. The answer arrives late and may miss half the relevant work.

With a Knowledge Agent, the query can return project names, relevant deliverables, recurring themes, approved proof points, and source documents. It can also distinguish between externally shareable case material and internal-only notes.

This changes more than search. It changes the firm’s ability to compound its experience.

Don’t automate a broken handoff

There is a temptation to buy an enterprise AI licence and begin asking people to use it. That may create isolated productivity gains, but it doesn’t create an operating model.

The stronger sequence is to map the workflow first.

For a proposal process, identify the trigger, the inputs, the approval points, the source material, the responsible roles, the system of record, and the output standard. Then determine which steps an agent can perform and where human review is required.

For research, decide which sources are acceptable, how citations should appear, what questions the brief must answer, and which claims require primary verification.

For knowledge, establish permissions. Client work has confidentiality obligations. A useful Knowledge Agent must respect workspace boundaries, access controls, document retention rules, and the difference between reusable intellectual property and client-specific content.

This is the practical work that clients need help doing too. Firms that build it internally develop a stronger delivery story. They don’t say, “We know AI is important.” They can say, “Here is the workflow, the control model, the review design, and the business measure we used.”

If you’re building that capability, Omni Ops is focused on agent-led business workflows rather than stand-alone prompts. You can also review Omni Advisory if the first need is shaping the operating model, governance, and implementation roadmap.

Put a dollar value on the manual work

A good business case doesn’t need inflated claims. It needs a conservative view of capacity and cost.

Take proposals. If your firm produces 18 substantial proposals a year and each consumes 25 hours of senior and manager time, that is 450 hours. An agent-supported process won’t remove every hour. It should not. But reducing drafting, searching, and formatting effort by 30 to 50 percent can return meaningful capacity to client work, relationship development, or better pursuit strategy.

Research compounds even faster. A firm with 10 to 20 active projects may have analysts repeating similar market and company research every month. The cost is not merely the initial research time. It is the review time, rework, inconsistency, and lost ability to reuse a good brief.

Knowledge debt is harder to see because it appears as interruption. A director asks a manager for an example. A manager asks an analyst to find a document. Three people spend an hour looking for something the firm already owns. Across a year, that becomes a real cost.

For firms in the $1 million to $25 million range, an $80K to $300K leakage band is a useful place to investigate. The exact number depends on utilisation, wage structure, proposal volume, and how much of the work is already standardised. The point is to calculate it from your own workflow, not borrow a headline number.

A 60-minute audit can make this concrete. Book a 60-min Omni Audit and we will identify the workflow with the best early economics, define what an agent should do, and outline the controls needed to deploy it. There is no slide deck to sit through. You leave with three practical outputs: a leakage view, a prioritised agent opportunity, and a clear next-step plan.

Use the first agent to build delivery credibility

Don’t try to turn your whole firm into an AI business in one quarter.

Choose one workflow where there is enough volume to matter, enough structure to be reliable, and enough pain that the team will adopt it. Proposal generation is often the best commercial starting point. Research is often the best delivery starting point. Knowledge retrieval can be the highest leverage once your documents and permissions are ready.

Set a baseline before deployment. Measure proposal preparation hours, time to first draft, research preparation time, review cycles, and the number of times staff need to request past work from colleagues. Then compare those figures after the workflow has been in use for several weeks.

That evidence becomes part of your client offer.

One trades-business owner in our network describes the value simply. He doesn’t want a vendor telling him what AI can do. He wants someone to take a frustrating process, make it work better, and show him the result. Corporate clients are moving toward the same expectation, even when the underlying work is more complex.

For a practical way to define that first workflow, download Deploy Your First Business Agent. It is a working checklist for choosing a process, setting inputs and review points, and avoiding the common mistake of deploying a tool before the job is clearly defined. You can also access the direct worksheet here.

The opportunity is not reselling AI licences

OpenAI’s enterprise revenue passing consumer revenue is a signal that AI purchasing has entered the corporate mainstream. Your clients are not waiting to be introduced to the category.

The consulting firms that win from this shift will help clients move from access to adoption. They will design the work, apply the domain knowledge, build controls, train teams, and prove the commercial result.

Start by doing the same inside your own business.

Look at the 20 to 40 hours being spent on major proposals. Look at the research being recreated across engagements. Look at the intellectual property locked in project folders. Those are not back-office inconveniences. They are a direct constraint on margin, utilisation, and the firm’s ability to show clients what implementation looks like.

See Omni for consulting firms to assess where the strongest first use case sits in your firm. If you want to turn that assessment into a specific operating plan, Book my Omni Audit.