The phrase "AI consulting" covers a widening gap. At one end, traditional management consultancies repackage AI as a practice area and deliver it with the same methods they have used for decades: interviews, slide decks, and a roadmap to another engagement. At the other, agentic AI consultancies build and deploy working systems, hand the firm operational capability, and compound the value over time.
The distinction is not semantic. It changes what buyers should look for, how engagements should be scoped, and what success looks like at the 90-day mark. This guide sets out the practical difference between the two, where the economics diverge, and how mid-market buyers should choose.
Traditional consulting produces a report that recommends change. Agentic consulting produces a working system that is the change.
What Traditional AI Consulting Looks Like
A traditional consulting engagement on AI typically follows a familiar shape. Weeks one through three are interviews, discovery workshops, and desk research. Weeks four and five are analysis and roadmap construction. Week six is a presentation, usually to the board, of a 40-60 slide deck containing the findings and recommendations. At the end of the engagement, the firm has a document.
The deliverable is intellectually rigorous. It identifies the right opportunities, it sequences them correctly, it highlights the risks. And then nothing happens. The firm now has to find a separate implementation partner, negotiate a separate contract, and rebuild the operational context from scratch for a team that was not in the room during the diagnostic. The implementation rate of traditional AI consulting roadmaps at mid-market firms is estimated below 40%.
The economics reflect the shape. Traditional AI consultancies charge partner-rate day fees for analyst-delivered work, in six-figure engagement tiers, against deliverables that are predominantly documents. Ongoing value depends entirely on what the client does with the deck.
What Agentic AI Consulting Looks Like
An agentic AI consulting engagement replaces the deck with a working system. The diagnostic still happens — it must — but it is shorter, it is embedded in the implementation phase, and it is validated against real operational data rather than against stakeholder opinions. The consultant builds or configures the AI agents that operate on the firm's actual workflows, integrates them with the firm's existing systems, and hands over a working capability.
The deliverables are qualitatively different. Where a traditional engagement produces a strategy document, an agentic engagement produces agents in production: a client intake triage agent handling inbound enquiries, a billing assistant drafting matter narratives, a research assistant producing first-draft memoranda. Each one is observable, measurable, and operationally embedded from day one.
Agentic consulting engagements also carry a different ongoing shape. A well-designed agent needs tuning, monitoring, and iterative improvement — not a retainer for "ongoing advisory" but an operational partnership for ongoing capability. The consultancy stays close to the system because the system continues producing value.
Where the Economics Diverge
The cost structures look superficially similar. Both types of engagement bill somewhere in the five-to-six-figure range for mid-market clients. The difference is in what the money buys.
Traditional consulting cost structure
Senior time for interviews and analysis: typically 60-70% of the engagement cost.
Junior time for research, formatting, and deck construction: 20-30%.
Deliverable production: 5-10%.
Implementation: outside scope.
Value is front-loaded at delivery of the deck. If the firm does not implement, the value is zero — and there is no contractual mechanism by which the consultancy is accountable for implementation outcomes.
Agentic consulting cost structure
Diagnostic and scoping: typically 15-25% of the engagement cost.
Build and integration of working systems: 50-60%.
Testing, change management, and handover: 15-20%.
Ongoing operational partnership: separate from the initial engagement.
Value is tied to production — if the system is not operating in the firm at engagement close, the engagement has failed. Accountability is structural, not rhetorical.
What Each Is Actually Good For
Traditional consulting is not obsolete. It remains the right tool for certain problems:
Board-level strategic alignment where the deliverable genuinely is a document that partners can debate and adopt.
Industry benchmarking and competitive positioning — work that does not produce operational capability but does produce informed decisions.
Regulatory or compliance assessment where the primary output is a written determination, not a system change.
Transaction support where a deck is the deliverable by definition (M&A due diligence, for example).
Agentic consulting is the right tool for different problems:
Operational transformation where the firm needs new capability, not a description of the capability it should build.
AI strategy where the strategy is meaningless without the agents that execute it.
Process redesign on revenue-touching workflows where the consultancy and the client are building the new workflow together.
Capability building where the engagement transfers not just knowledge but operational artefacts.
How Buyers Should Choose
The choice depends on what the firm actually needs at this moment.
Choose traditional consulting when
The deliverable you need is a document, and the document itself is the goal.
You have, or will imminently have, a separate implementation partner already identified.
The problem is primarily strategic alignment, not operational capability.
Your partnership has a strong preference for deliberation before action and would not absorb operational changes mid-engagement.
Choose agentic AI consulting when
You have diagnosed the problem already (or can self-diagnose from a short engagement) and need working capability next.
Your firm has tried and failed with a traditional AI consulting engagement that produced a deck nobody implemented.
You want the consulting partner to be accountable for operational outcomes, not just analytical quality.
Your fee-earners will be using the systems directly and you need change-management built into the build.
The Hybrid Case
Some engagements genuinely need both. A firm considering a twelve-month programme of AI adoption across a dozen initiatives may need a traditional strategic piece to align the partnership on sequencing — followed by agentic engagements that build the specific systems. The right structure is usually a short strategic piece (days, not weeks) followed by agentic implementation.
What does not work is stretching a traditional engagement to simulate implementation. Adding a fifth week to a four-week deck engagement does not make it an agentic engagement. Either the consultancy has the capability to build working systems or it does not; and the ones that do not are increasingly obvious in the market.
The UK Mid-Market Reality
For UK mid-market firms specifically — 5 to 150 fee-earners for law firms, equivalent scale for other professional services — the agentic model is almost always the right starting point. The reasons are practical:
Budget. Mid-market firms cannot afford the traditional six-figure deck engagement and then a separate six-figure implementation. A single agentic engagement delivers the operational outcome within the budget that would otherwise fund only the deck.
Attention. Mid-market partner bandwidth is scarce. A traditional engagement that requires the firm to run a separate implementation project after the diagnostic has a high probability of stalling for that reason alone.
Proof. Mid-market firms benefit disproportionately from seeing working AI in their own environment early. A deck cannot demonstrate what an agent operating in the firm can. Belief follows evidence.
Competitive dynamics. Mid-market firms are competing against other mid-market firms and against ABS entrants. The ones that deploy agents first are the ones that change their operating economics first. A deck does not change operating economics.
Red Flags in Either Model
Whichever model you pick, some signals should cause the firm to walk away:
No reference clients at comparable size and sector.
Unwillingness to name the specific AI platforms or models that will be used.
No measurement framework for success — if the consultancy cannot tell you how outcomes will be measured, they do not have a plan to produce outcomes.
Scope that is ambiguous by design. "We'll figure it out together" is a billing strategy, not an engagement strategy.
Locked deliverables. Findings you cannot show to another partner, opportunities you cannot pursue without the consultancy's continued involvement — these are captures, not consulting engagements.
The Strategic Choice
The real distinction between traditional and agentic AI consulting is not technological. It is about where value is created and where accountability sits. Traditional consulting creates value in the analytical process and holds accountability at the document. Agentic consulting creates value in the operating system and holds accountability at the outcome.
For most mid-market firms choosing their first AI consulting engagement, the agentic model is the better commercial fit. For strategic work that genuinely needs a document, the traditional model still has a role — but it should be scoped to the document, priced accordingly, and followed by an agentic engagement for the implementation.
The question is not "traditional or agentic?" as a general preference. It is "what outcome do we need, and which engagement shape produces that outcome most efficiently?" — asked for each engagement, not once and for all.