AI consulting: What to expect before you hire

When inquiries sit unanswered or staff copy the same information between systems, buying advice is not the end goal. Getting that work done is.

AI consulting should connect the business problem to a specific change in how work happens. Before hiring a firm, ask what you will receive, what decisions those deliverables let you make, and who will get the system into everyday use.

The first question is not “Which model should we use?” It is “Do we need to work out what to build, or does the solution already exist?”

What exactly does an AI consultant do?

An AI consultant examines a business problem, decides whether AI is appropriate and defines how to implement it. Depending on the engagement, they may also build, connect, test and operate the resulting system.

Those are different responsibilities. Do not assume an AI consulting service includes all of them.

Consider an inbound sales inquiry. A prospect asks about a service, but the team still needs to:

• Answer from the current service and pricing information.
• Collect missing details.
• Decide whether the request fits.
• Find an appropriate appointment.
• Book it and update the CRM.
• Refer unusual requirements to a person.

A consultant should identify where that chain breaks and what would fix it. The answer might involve an agent, a simple rule, a clearer policy or a change in who owns the request.

If you have not chosen a process yet, use our repeated business work self-assessment to find a sensible starting point.

What should an AI consulting engagement produce?

An AI consulting engagement should leave you with decisions and working evidence, not just recommendations. Each deliverable should answer a question that would otherwise block implementation.

For a custom workflow, ask for the following.

Deliverable	What it should contain	Decision it enables
Current workflow map	Trigger, people, systems, decisions, exceptions and handoffs	Where does the work actually get stuck?
Measured baseline	Volume, handling time, error rate and current cost	Is this problem worth solving?
Scope and responsibility split	Agent tasks, rule-based steps and human decisions	What will change, and what will not?
Implementation design	Connections, knowledge sources, access limits and failure handling	Can this run within our systems and policies?
Acceptance tests	Representative past cases and explicit pass criteria	What counts as good enough to go live?
Rollout plan	Shadow testing, limited release, approvals and rollback	How will we contain mistakes?
Operating plan	Monitoring, exception ownership, maintenance and reporting	Who keeps it working after launch?

These do not all need to arrive in one large document. They do need to be explicit.

For the inquiry example, “automate lead qualification” is not sufficient scope. A usable specification names the qualification questions, booking rules, permitted promises and situations requiring a person.

A useful acceptance test might ask: when a prospect requests a service outside the approved offering, does the system avoid making a promise and pass the request to the right person?

For the mechanics of dividing the work, see our guide to redesigning one workflow with AI automation.

When is a discovery audit necessary?

A discovery audit is useful when you cannot yet define the workflow, the right intervention or the conditions for success. It should resolve uncertainty before you commit to a build.

Discovery is justified when:

• Several departments handle the same request differently.
• Nobody has a reliable view of current volume or handling time.
• Important decisions depend on undocumented judgment.
• You do not know whether the problem needs AI or straightforward automation.
• Access, data quality or policy constraints could change the design.
• There are several opportunities, but no agreed priority.

The audit should end with a bounded first project and a success target. A list of possible AI use cases is not enough.

Ask before signing: “What will we know at the end that we do not know today?”

If the answer is vague, the discovery scope probably is too.

When can existing agents make an audit unnecessary?

A separate discovery audit is unnecessary when an existing agent already covers the workflow and the remaining work is confirming fit, connecting systems and setting business rules. You still need testing and configuration. You do not necessarily need a separate engagement to discover what to build.

For example, suppose the requirement is already clear: answer approved pricing questions, gather qualification details, book a meeting and save the conversation to the CRM.

At Dapto, those tasks are covered by our existing Customer Service Agents. We can start with those agents and adapt them to the business rather than requiring an AI Opportunity Audit first.

The fit checks still matter:

• Are the approved answers current?
• Which information must be collected before booking?
• Which calendar and CRM should be used?
• What can the agent complete without approval?
• Which requests must reach a person?

Our agents use a shared knowledge source called the Brand Core for the business's information, policies and voice. Requests beyond the agreed limits go to a person with a summary and the conversation.

Skipping an audit does not mean skipping controls. It means the core workflow is already covered, so the work begins with adapting and testing it.

An unusual process that the existing agents do not cover is different. That is where our custom work begins with an audit.

Should you hire an AI consulting firm or an AI automation agency?

Choose by the work included, not the label. An AI consulting company, AI consulting agency and AI automation agency may offer overlapping services, so the proposal matters more than the name.

Type of help	Best fit	What to verify
AI strategy consulting	You need priorities, investment decisions or a broader operating plan	Does the strategy lead to scoped projects with owners?
AI automation consulting	You know the recurring work that needs attention	Can the provider connect the whole process, including exceptions?
AI implementation consulting	You have a direction but need it built and deployed	Are testing, rollout and operating responsibility included?
Existing-agent customisation	A ready-built agent covers the core job	What needs adapting, and what falls outside its capabilities?
Internal build with specialist advice	Your team will own engineering and operation	Who owns each technical decision and production issue?

An AI automation consultant may be right for a defined reporting or intake process. An AI implementation consultant may be more useful when a prototype exists but permissions, integrations and testing remain unresolved.

If you are comparing delivery models rather than individual firms, our done-for-you AI automation services guide goes deeper into project, managed-service and self-serve options.

How should you evaluate an AI consulting company?

Evaluate an AI consulting company against one real workflow. Ask each provider to explain what happens from the incoming request through to the completed action and recorded outcome.

Use the same questions across your shortlist:

1. What would you reuse, and what must be built? Ask for a clear boundary between existing capability and custom development.
2. What stays with our team? Look for named approval points and exception owners, not a vague promise of human oversight.
3. What happens when information is missing or contradictory? The answer should include a safe stopping point and a route to a person.
4. How will you test it? Ask for tests using representative cases, including awkward ones.
5. What access does it need? Require an explanation of what the agent can read, change and send.
6. How will we know it is working? Measures should reflect completed work, errors and human involvement, not just message counts.
7. Who supports it after launch? Clarify monitoring, corrections, changes and incident responsibility.
8. What evidence supports your claims? Separate demonstrations and illustrative examples from verified client results.

Be cautious if a proposal promises a transformation without naming the first workflow. Also question a proposal that requires replacing core software without explaining why.

How much does it cost to hire an AI consultant?

The cost depends on the scope and responsibility you are buying. Advice, an integrated production system and ongoing operation are not equivalent purchases.

Ask for a breakdown covering:

• Discovery, if needed.
• Configuration or custom development.
• Connections to existing systems.
• Testing and rollout.
• Software and usage charges.
• Monitoring and ongoing changes.
• Work your own team must contribute.

Compare proposals against the same endpoint. A lower build quote may exclude testing or support that another proposal includes.

At Dapto, accelerator customisation has no standard list price. Custom work is scoped after the AI Opportunity Audit, with pricing based on the workflows, systems and agents involved.

To judge the value, start with your own workload rather than a provider's savings claim. Our AI automation ROI guide explains how to build that baseline.

What does this process look like with Dapto?

We distinguish between adapting an existing agent and designing a workflow that needs custom work. Only the second path starts with an AI Opportunity Audit.

For custom engagements, the outputs are:

1. AI Opportunity Map. We examine the work and its baseline, separate agent tasks from rules and human decisions, then agree the first build and success target.
2. AI Workflow Architecture. We define responsibilities, tools and permissions, then build tests from past cases with the business's experts.
3. Production AI System. We connect the existing systems with limited access, test beside the team, then release a small share of live work before expanding. Sensitive actions retain approval, with a rollback plan.
4. Managed AI Operations. We monitor runs, review exceptions, re-test changes and report monthly against the baseline.

The systems remain in place. The work is connecting and coordinating them, not requiring a migration.

For a covered workflow, we start with the existing agents instead. A business that wants to test Customer Service Agents independently can also use our seven-day free trial, with same-day setup and no developer required.

The next step is a short call about one process: how requests arrive, how much work they create, what the team does today and which decisions should remain with people.

FAQ

Is AI consulting worth it for small businesses?

AI consulting can be worthwhile for a small business when a specific recurring process consumes enough time or loses enough opportunities to justify the work. Start with one bounded problem rather than a company-wide strategy. If an existing agent covers that problem, customisation or a self-serve trial may be more appropriate than a discovery engagement.

What are the top AI consulting companies in the USA?

There is no single ranking that identifies the right AI consulting firm for every business. Compare providers against your workflow, systems, controls and required support. A broad strategy engagement and a narrowly scoped implementation need different strengths.

What is the difference between AI strategy consulting and AI implementation consulting?

AI strategy consulting determines where to invest and what to prioritise. AI implementation consulting turns an agreed opportunity into a tested system connected to business operations. Confirm whether your engagement covers one or both, and who owns the handoff.

Do we need an audit before buying AI agents?

You need discovery when important questions about the workflow or solution remain unresolved. You do not automatically need a separate audit when an existing agent already covers the job. Fit checks, business rules, permissions and acceptance testing are still necessary.

How much money do AI consultants make?

Consultant earnings are a career question, not a reliable guide to what your engagement should cost. Employee salary, independent billing rates and a firm's project fees measure different things. As a buyer, compare the defined scope, deliverables and ongoing responsibilities.

What should we prepare before speaking to an AI automation consultant?

Bring a few representative requests, the systems involved, the rules staff follow and examples that require judgment. Add whatever you know about volume, handling time and errors. You do not need to choose a model or know which agent you need before the conversation.