AI Consultants vs an AI Audit vs Doing It Yourself: Which Should You Choose?
Use this article with AI
Open a buyer-ready brief in your preferred assistant.
ChatGPT, Claude, Perplexity and Grok open with the prompt attached. Gemini copies the prompt to your clipboard first.
Mitori
Article assistant
Ask about this article
I can help you understand the key arguments, how they apply to your organisation, and what questions to ask next.
Powered by Mitori · Scoped to this article
Do it yourself when the workflow is small, simple and owned by one team. Hire an AI consultant when you need strategy, change management or a framework leadership already trusts. Choose an observation-based AI audit when you need to know which workflows to automate first, based on how work actually happens, and want one partner through build and monitoring.
Mitori is an operational intelligence platform that observes how work actually happens across your people, tools and handoffs, then tells you what to automate first, in what order and under what controls — and builds and monitors the AI agents that do it.
What are your three options for getting AI into the business?
Most companies choose between building it themselves, hiring an AI consultant or strategy firm, or commissioning an AI audit that studies how work actually happens before anything is built. All three can work. They fail in different ways, so the right choice depends on what you are missing.
Doing it yourself means a capable internal person or team picks a use case, buys or configures tools, and ships. Hiring a consultant means bringing in an outside team to run discovery, usually through interviews and workshops, and produce a strategy and roadmap. An observation-based AI audit means watching real work across the tools your teams already use, with consent, then ranking and sequencing what to automate from that evidence.
The stakes are real. According to MIT NANDA's "The GenAI Divide: State of AI in Business 2025" report, as covered by Fortune, about 95% of generative AI pilots deliver little to no measurable P&L impact. The researchers blamed a learning gap and flawed enterprise integration, not model quality. The same research found that buying from specialized vendors and building partnerships succeeded about 67% of the time, while internal builds succeeded only about one-third as often.
That finding does not mean DIY is wrong. It means integration into real workflows is the hard part, and you should pick the path most likely to get that part right for your situation.
How do DIY, consultants and an AI audit compare?
The biggest difference is the evidence each path relies on. That evidence decides how much you can trust the roadmap, and it shapes what you are left holding when the engagement ends.
Interviews are not worthless. They capture intent, politics and constraints that no log will show. But people describe how work is supposed to happen, and they underestimate the small, repeated tasks that eat the day. Observation fills that gap: it records the handoffs, rework and workarounds that rarely come up in a workshop.
A deck is also not worthless. A clear strategy document aligns leadership. The question is whether anyone is accountable for turning it into running systems.
| Path | Where the evidence comes from | What you have at the end |
|---|---|---|
| Doing it yourself | Your team's own knowledge of the work, plus whatever tool vendors show in demos | A working tool or pilot, owned in-house, often without a wider roadmap or a measured baseline |
| AI consultant or strategy firm | Stakeholder interviews, workshops, surveys and the firm's frameworks and benchmarks | A strategy and roadmap, usually as a deck, sometimes with implementation offered as a separate engagement |
| Observation-based AI audit (Mitori) | 30, 60 or 90 days observing real work across email, spreadsheets, chat and business systems, with consent | A sequenced, confidence-scored roadmap, then agents built and monitored against the observed baseline |
Path
Doing it yourself
Where the evidence comes from
Your team's own knowledge of the work, plus whatever tool vendors show in demos
What you have at the end
A working tool or pilot, owned in-house, often without a wider roadmap or a measured baseline
Path
AI consultant or strategy firm
Where the evidence comes from
Stakeholder interviews, workshops, surveys and the firm's frameworks and benchmarks
What you have at the end
A strategy and roadmap, usually as a deck, sometimes with implementation offered as a separate engagement
Path
Observation-based AI audit (Mitori)
Where the evidence comes from
30, 60 or 90 days observing real work across email, spreadsheets, chat and business systems, with consent
What you have at the end
A sequenced, confidence-scored roadmap, then agents built and monitored against the observed baseline
What drives the speed, cost and ownership of each path?
Cost is driven less by day rates or license fees than by how long it takes to reach a decision you trust, and by who carries the outcome after go-live. Each path moves that burden somewhere different.
DIY looks cheapest on paper, and for a contained problem it often is. The hidden cost shows up when a team automates a workflow that turns out to be low value, or that breaks at the first exception nobody mapped.
Consulting costs are easier to see up front, but the risk is the gap between recommendation and production. If the firm that wrote the roadmap is not the one building and running it, assumptions get lost in the handoff.
With an observation-based audit, the observation window is the main time cost. What you get in exchange is a baseline: a measured picture of how long work takes today, which every later ROI claim can be checked against. For engagement structure, see the pricing page rather than any figure quoted in an article.
| Path | Speed and cost drivers | Who owns the outcome |
|---|---|---|
| Doing it yourself | Fast to start. Cost is mostly internal time, tool subscriptions and the opportunity cost of picking the wrong workflow | You do, fully, including maintenance, failures and governance |
| AI consultant or strategy firm | Weeks of discovery and workshops. Cost scales with team size, duration and how much implementation is added | The firm owns the recommendation. You or a separate implementer usually own delivery and results |
| Observation-based AI audit (Mitori) | Audit length (30, 60 or 90 days) and scope set the timeline. Cost scales with scope and the number of agents built and monitored | The same partner carries the work from audit through build and control, with your team approving each wave |
Path
Doing it yourself
Speed and cost drivers
Fast to start. Cost is mostly internal time, tool subscriptions and the opportunity cost of picking the wrong workflow
Who owns the outcome
You do, fully, including maintenance, failures and governance
Path
AI consultant or strategy firm
Speed and cost drivers
Weeks of discovery and workshops. Cost scales with team size, duration and how much implementation is added
Who owns the outcome
The firm owns the recommendation. You or a separate implementer usually own delivery and results
Path
Observation-based AI audit (Mitori)
Speed and cost drivers
Audit length (30, 60 or 90 days) and scope set the timeline. Cost scales with scope and the number of agents built and monitored
Who owns the outcome
The same partner carries the work from audit through build and control, with your team approving each wave
Read next:How Mitori prices the audit →How to measure AI ROI before you buy →
When does each path make sense?
Match the path to the problem. Simple and contained points to DIY. Organizational and strategic points to a consultant. Cross-team workflows where you do not yet know what to automate first point to an observation-based audit.
Mitori's team helped scale a healthcare staffing company, SnapNurse, to $1.3B in revenue, and Mitori's agents already run in healthcare operations, including IVF clinics and healthcare staffing agencies. That background shapes the bias toward observing operations before recommending anything.
- DIY fits when one team owns the workflow end to end, the steps are stable and well understood, the data is not sensitive, and someone internal has time to build and maintain it. A small business automating invoice reminders or meeting notes rarely needs outside help.
- An AI consultant fits when the hard problem is organizational: setting an AI policy, aligning a leadership team, restructuring roles, preparing a board strategy or running change management across many functions. Good firms bring frameworks, benchmarks and facilitation that internal teams lack.
- An observation-based AI audit fits when work spans several tools, teams and handoffs, when leaders disagree on where AI should go first, or when finance wants a baseline before approving spend. Mitori is built for this case: it observes the work, scores each automation on savings, effort and confidence, sequences it into deployment waves with approvals and escalation, then builds and monitors the agents.
- Combinations are common. A consultant can own policy and change while an audit supplies the workflow evidence, or an internal team can build the simple wins while an audit sequences the cross-team work.
Read next:How to deploy AI in your business →Where should a company start with AI? →See how the Mitori audit works →
What should you ask any AI partner before you sign?
Ask questions that expose the evidence, the baseline, the ownership and the controls. A credible partner, whether consultant, vendor or auditor, will answer each one specifically.
The answers matter less for their polish than for their specificity. A partner who cannot say how they will baseline today's work cannot credibly tell you what AI will save.
- 1Where will your recommendations come from: interviews, system data, observation of real work, or your own templates?
- 2How will you measure how the work runs today, so we can check any savings claim against a baseline?
- 3What exactly do we receive at the end: a document, a ranked roadmap, working systems, or all three?
- 4Who builds what you recommend, and who is accountable if it underperforms after go-live?
- 5Which workflows would you tell us not to automate yet, and why?
- 6What is the governance plan: approvals, escalation paths, human review, logging and how drift is detected?
- 7How do you handle consent, data access and privacy during discovery?
- 8If we part ways, what do we keep: the evidence, the roadmap, the agents, the logs?
What red flags should you watch for?
Watch for relabeled products sold as agents, ROI numbers with no baseline behind them, and roadmaps with no governance plan. Any one of these signals a project likely to stall.
- Agent washing. Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027 due to escalating costs, unclear business value or inadequate risk controls, according to reporting on its June 2025 forecast. Gartner also warned that many vendors relabel existing assistants, chatbots and RPA as agents, and estimated only about 130 of thousands of agentic AI vendors are real. Ask what the system decides on its own, and what it escalates.
- ROI promises without a baseline. A percentage saving quoted before anyone has measured the current workflow is a sales number, not a forecast. Ask how the figure was built and what it assumes.
- No governance plan. If approvals, escalation, logging and human review are deferred until after launch, they usually arrive as rework. Frameworks such as the voluntary NIST AI Risk Management Framework, organized around Govern, Map, Measure and Manage, give you a checklist to hold any partner to.
- A deck with no path to production. Strategy without an owner for build and operation tends to become shelfware.
- Pilots designed to impress, not to scale. A demo on clean sample data says little about exceptions, handoffs and the systems your team actually uses.
Read next:Why AI pilots fail →AI agents vs automation vs chatbots →How Mitori handles governance →
How do you decide which path to take?
Start by naming what you are missing: build capacity, organizational alignment, or evidence about where AI belongs. That gap usually points to the path.
If you are unsure where you stand, the readiness assessment gives a quick read on whether your organization has the evidence, ownership and controls to move. If cross-team workflows are the problem, Mitori's Audit → Roadmap → Build → Control model is designed to take you from observed work to governed agents with one accountable partner.
- 1List the two or three workflows people most want to automate, and note how many teams and tools each one crosses.
- 2If a workflow sits inside one team and one tool, try DIY first with a clear owner and a simple before-and-after measure.
- 3If the blocker is leadership alignment, policy or change management, bring in a consultant with a defined scope.
- 4If the work crosses teams and nobody can agree what goes first, commission an observation-based audit to produce a baseline and a sequenced roadmap.
- 5Before signing anything, run the questions above and check for the red flags.
Read next:Take the readiness assessment →See the audit model →Review proof of the model →
Frequently asked questions
Not always. If the workflow is simple, owned by one team and uses low-risk data, an internal owner can often pilot it with off-the-shelf tools. A consultant earns their fee when the challenge is strategy, policy or change across many teams. If the challenge is knowing which workflows to automate first, an observation-based audit gives you evidence rather than opinions.
An AI readiness audit assesses whether your workflows, data, ownership and controls are ready for automation, and where AI should go first. Many audits rely on interviews and questionnaires. An observation-based audit, like Mitori's, watches real work across the tools teams already use, with consent, then produces a sequenced, confidence-scored roadmap with governance checkpoints.
It depends on scope and method. Questionnaire-based assessments can take days, while workshop-led discovery often runs several weeks. Mitori's audit runs for 30, 60 or 90 days of observation, because recurring workflows, exceptions and handoffs need time to show up in the evidence. Longer windows suit larger scopes and more variable work.
On direct spend, often yes. The real cost of DIY is internal time, maintenance and the risk of automating the wrong workflow. MIT NANDA's 2025 research found internal builds succeeded about one-third as often as purchased tools and partnerships. In-house works well for contained problems; cross-team work usually benefits from outside evidence and delivery support.
At minimum: a baseline of how the work runs today, a ranked and sequenced list of what to automate, a governance plan covering approvals, escalation and logging, and a clear owner for build and operation. Be wary of partners who deliver only a deck or only a tool, with nothing connecting the recommendation to measured results.
Next step
Book an audit scoping call
Weighing a consultant, an in-house build or an audit? Walk through your workflows with us and see what an observation-based audit would cover.
Sources
Related reading
AI Deployment Guide
How to Deploy AI in Your Business: A Step-by-Step Guide (2026)
How to deploy AI in your business in 7 steps: observe real work, pick workflows by evidence, set governance, build agents with approvals, then scale in waves.
Getting Started
Where Should a Company Start With AI? How to Find the Workflows Worth Automating
Where to start with AI in business: skip the brainstorm. Score real workflows on volume, repetition, handoffs, error cost, data and risk, then start there.
AI Strategy
Why AI Pilots Fail (and How to Get to Production)
Why AI pilots fail: the research points to workflow fit, missing baselines and late governance, not model quality. Seven failure modes and the fix for each.
AI ROI
How to Measure AI ROI Before You Buy Any Tools
How to measure AI ROI before you buy: capture an observed baseline, model savings as ranges with confidence, price effort and risk, then track realized value.
Category Comparison
Process Mining vs Operational Intelligence for Enterprise AI
Process mining is useful where system logs are complete. Operational intelligence matters when enterprises need to reconstruct cross-tool work and decide what to automate next.