By Dean McCoubrey Chief AI Strategist
Before adopting AI, a business should define what AI is meant to improve: decision quality, commercial intelligence, team capability, brand distinctiveness, or operational speed. The mistake is starting with tools before clarifying the outcomes AI should support. Without that clarity, AI adoption creates more content, more automation, and more activity, but not better decisions, stronger brands, or more commercial advantage.
Most businesses ask what AI can do. The more useful question is what AI should help the business become better at.
That is a small shift in framing. It turns out to make a large difference in results.
Key point: AI adoption is not primarily a technology decision. It is a decision about what kind of business you want to become, and how human judgement, commercial intelligence, and organisational capability should be strengthened in the process.
This article works through that decision in a practical sequence.
Why most AI adoption starts in the wrong place
The pressure to adopt AI is real. Boards want it on the agenda. Teams are already experimenting. Competitors are announcing strategies. In that environment, the easiest move is to pick a platform, run a pilot, measure hours saved, and call it progress.
That sequence feels productive. It rarely is.
MIT’s research into the state of AI in business found that most AI systems become “static science projects” that fail to adapt, improve, or integrate into production workflows. The dominant pattern, as their analysis describes it, is not that AI does not work. It is that organisations cannot turn prototypes into operating systems that change workflows and deliver measurable ROI.
The numbers support this. Enterprise AI implementation failure rates are commonly cited in the 70 to 85 percent range, driven not by model capability limits but by organisational, data, and integration failures. According to Deloitte’s State of AI in the Enterprise 2026, readiness is now the bottleneck, not access to AI.
Tool-first adoption produces a predictable set of problems:
- Tool sprawl: multiple platforms, no shared standard, inconsistent outputs
- Inconsistent quality: AI produces content and decisions without clear review points
- Weak governance: nobody owns the AI inventory or the standards it operates within
- Unclear ROI: activity metrics replace commercial outcomes as the measure of success
- Amplified weaknesses: poor positioning, weak data, and disconnected workflows scale rather than improve
AI does not fix unclear strategy. It scales it. That is the risk most businesses underestimate before they begin.
The five questions to answer before adopting AI
Before choosing a tool, platform, or workflow, every business should answer five questions. Not as a compliance exercise. As a strategic sequence.
Each question changes the kind of AI system a business needs, the governance model required, and the workflows that will need to change. Skip them and you are not adopting AI strategically. You are buying capability without knowing what it is for.
The questions are:
- What should AI improve?
- Which human capabilities must stay in control?
- What decisions will AI support?
- Is the operating model ready?
- How will you measure whether AI is working?
Work through them in order. The answers to the first question shape the answers to all the rest.
1. What should AI improve?
This is the first and most important question, and the one most businesses skip. The default answer is “productivity.” That is almost always the least strategic starting point.
AI can improve very different things, and each requires a different system, a different review model, and a different definition of success.
| Business outcome | What AI should do | What changes |
|---|---|---|
| Speed | Accelerate content, research, and production workflows | Volume and turnaround time |
| Decision quality | Surface patterns, signals, and insights faster | Commercial judgement and strategic clarity |
| Brand consistency | Maintain tone, standards, and quality at scale | Output quality and distinctiveness |
| Customer intelligence | Identify needs, segments, and behaviour patterns | Targeting and relevance |
| Team capability | Remove low-value tasks, create space for strategic work | How people spend their time |
| Commercial performance | Improve pipeline quality, conversion, and growth | Revenue and competitive advantage |
A business trying to produce content faster needs a different AI setup from one trying to improve strategic decision-making or commercial intelligence. These are not the same problem.
If you cannot say what AI is meant to improve, you are not ready to choose the tool.
That is not a reason to delay. It is a reason to spend one focused session defining the outcome before spending a budget on the platform.
2. Which human capabilities must stay in control?
Quality drops when AI is asked to make decisions that require human judgement rather than support execution. This is not a technology limitation. It is a design failure: the business did not decide, before adoption, which capabilities must remain human.
According to Bynder’s research, 52% of consumers reduce engagement when they suspect content is AI-generated. Salesforce data shows only 4% of marketers highly trust AI-generated content without human oversight. The risk is not that AI produces bad work. It is that businesses let AI make decisions that only humans can make well.
At Humaine, we tend to see the same pattern when AI adoption goes wrong: the business let AI make decisions that only humans can make well. Seven capabilities in particular need to stay active:
| Capability | Why it must stay human |
|---|---|
| Imagination | Determines whether AI expands possibilities or repeats familiar patterns |
| Curiosity | Shapes the quality of questions AI is asked to answer |
| Creativity | Decides whether output is distinctive or merely competent |
| Empathy | Keeps outputs connected to real human needs and context |
| Judgement | Decides what should be acted on, published, or pursued |
| Taste | Separates acceptable output from work that is genuinely distinctive |
| Meaning-making | Connects AI outputs to the larger brand or business narrative |
These are not soft skills to protect from efficiency pressure. In a world where AI makes competent execution cheap and abundant, they are where competitive advantage actually lives.
AI should create more space for them, not replace them. That is the standard worth holding any adoption plan to. For a deeper view of how this works in practice, see our thinking on how to use AI without losing brand quality.
3. What decisions will AI support?
Most AI adoption conversations focus on outputs: content produced, tasks completed, hours saved. The more useful question is which commercial judgements AI can improve.
That is where the lasting advantage tends to sit. Not in volume, but in the quality of the choices a business makes.
The highest-value AI use cases improve decisions, not just throughput. Examples worth considering before adoption:
- Which market signals actually matter for our category right now?
- Which customer segments should we prioritise this quarter?
- Which content themes are worth investing in versus abandoning?
- Which campaigns are worth scaling, and which are burning budget?
- Which sales opportunities deserve focus, and which are unlikely to convert?
- Where are we missing patterns in our commercial data?
These are not workflow questions. They are commercial intelligence questions. And they require a different kind of AI setup: better data, clearer decision rights, and human judgement in the loop at every meaningful point.
The businesses getting the most from AI are not simply producing more. They are deciding better.
That distinction matters enormously. Speed without better judgement is just more noise. For a fuller view of how commercial intelligence differs from standard market research, see our piece on commercial intelligence versus market research.
4. Is the operating model ready?
AI adoption is not a tool decision. It is an organisational design decision. It changes workflows, roles, review points, governance, data movement, and accountability. Businesses that treat it as the former and ignore the latter end up with capable tools and broken processes.
Deloitte’s 2026 research found that only 20% of companies have a mature governance model for autonomous AI agents, even as adoption accelerates. As the Partnership on AI argues, the strongest approach is to treat governance as an operating system, not a compliance layer.
Not every business needs full AI-first transformation immediately. But every business should know which adoption mode it is choosing, because each carries different risks and requirements.
| Adoption mode | What it means | Key risk |
|---|---|---|
| Tool adoption | Individuals use AI for isolated tasks | Inconsistency and no shared standard |
| Workflow integration | AI is embedded into repeatable processes | Quality depends entirely on review design |
| Decision augmentation | AI improves judgement and commercial choices | Requires stronger data and governance |
| Operating model redesign | AI becomes part of how the business creates value | Requires leadership commitment and change management |
Most businesses currently sit between tool adoption and workflow integration. That is fine as a starting point. The risk is staying there indefinitely, adding tools without ever redesigning the workflows that determine whether those tools create value.
The question to answer before adoption is not “which mode should we reach eventually?” It is “which mode are we genuinely ready to operate in now?” For more on the difference between AI-first and AI-augmented approaches, see our thinking on AI-first versus AI-augmented marketing.
5. How will you measure whether AI is working?
Most AI projects are measured by activity: hours saved, content produced, tasks automated, cost reduced. Those are useful signals. They are also incomplete.
A business can improve all of them while quietly weakening the capabilities that create long-term commercial advantage. Getting faster and getting better are not the same thing.
A more useful AI scorecard measures outcomes across multiple dimensions:
- Decision quality: are commercial judgements faster and more accurate?
- Speed to insight: is the business acting on better information, sooner?
- Brand distinctiveness: is AI-assisted output raising or diluting the standard?
- Pipeline quality: are AI-supported processes improving conversion and revenue?
- Team capability: is AI creating more space for strategic and creative work?
- Reusable IP: is the business building assets that compound over time?
- Commercial performance: is growth improving as a direct result?
Deloitte’s research shows that AI leaders achieve an average return of 3.7x per dollar spent on generative AI. But only 5% of generative AI pilots deliver sustained value at scale. The gap between those two numbers is almost entirely explained by measurement: leaders track commercial outcomes, not just activity.
If your current AI success metrics are limited to efficiency, you are measuring the wrong things. For a practical view of how to connect AI investment to commercial outcomes, see our marketing maturity model assessment and our guide to proving marketing revenue.
A simple AI readiness diagnostic
Before choosing a platform or scaling a pilot, run through this short diagnostic. Answer yes or no to each question.
| # | Question | Yes / No |
|---|---|---|
| 1 | Do we know which business outcome AI is meant to improve? | |
| 2 | Have we defined where human judgement must remain in control? | |
| 3 | Do we know which workflows AI will change first? | |
| 4 | Do we have clear review points for AI-generated work? | |
| 5 | Do we have enough data quality to trust AI-supported decisions? | |
| 6 | Have we agreed who owns AI governance? | |
| 7 | Do we know how AI will affect brand quality? | |
| 8 | Have we identified the teams most likely to resist adoption? | |
| 9 | Do we know how we will measure success beyond time saved? | |
| 10 | Have we decided whether AI is a tool, workflow layer, or operating model shift? |
What your score means
| Score | Readiness level | Recommended next step |
|---|---|---|
| 0 to 3 | Not yet ready | Start with strategy, governance, and outcome definition |
| 4 to 6 | Partially ready | Pilot AI in one workflow with clear review points and measurement |
| 7 to 8 | Ready for structured adoption | Build repeatable workflows and a commercial measurement model |
| 9 to 10 | Ready for strategic integration | Move from tool adoption to operating model design |
A low score is not a reason to delay AI indefinitely. It is a reason to do the strategy work first, so that when you do adopt, you are building on a clear foundation rather than experimenting on top of an unclear one.
The businesses that struggle most with AI adoption are not those that moved too slowly. They are those that moved quickly without answering these questions, and then spent significant time and budget unwinding the consequences.
Deloitte’s research confirms this pattern: governance maturity remains low across most organisations even as adoption pressure increases. The gap between those who invest in readiness and those who do not is becoming one of the clearest dividing lines in enterprise AI performance.
What businesses should not do before adopting AI
Sometimes the clearest way to describe the right path is to name the wrong ones.
Do not start by buying tools. Tools chosen before outcomes almost always create sprawl. You end up with multiple platforms doing overlapping things, no shared standard, and a growing list of subscriptions that nobody is accountable for.
Do not automate broken workflows. AI scales process design, good and bad. If a workflow is unclear, inconsistent, or poorly owned before AI, it will be faster and more inconsistent after. Fix the process first.
Do not measure only hours saved. Efficiency is a useful signal, but it is not a growth strategy. A business can save thousands of hours while producing lower-quality work, weaker brand assets, and less distinctive commercial output. As one researcher put it plainly: “Productivity gains are meaningless if quality, trust, or risk deteriorate.”
Do not let AI define the brand voice. Brand distinctiveness depends on human judgement, taste, and meaning-making. AI can support execution and variation, but it should not decide what the brand believes or how it sounds. That decision must remain human.
Do not treat adoption as an IT project alone. AI changes how commercial decisions are made, how brand quality is maintained, and how value is created. Ownership must include commercial, marketing, and operational leadership, not just technology teams.
AI adoption is not an IT project. It is a capability, quality, and operating model decision. Treating it as anything less is how businesses end up with impressive tools and disappointing results.
The right first step
The right first step in AI adoption is not selecting a platform. It is getting clear on what AI is actually meant to improve, and whether the business is organised to use it well.
At Humaine, we work through that with B2B leaders as a structured review. It covers eight things:
- Business outcomes: what AI is specifically meant to improve, defined in commercial terms
- Human capabilities: where human judgement, creativity, and taste must remain in control
- Brand quality risks: how AI adoption could affect distinctiveness, trust, and standards
- Workflow priorities: which processes are ready for AI integration and which need redesign first
- Governance: who owns AI decisions, standards, and the inventory of systems in use
- Measurement: how success will be defined beyond efficiency metrics
- Commercial intelligence opportunities: where AI can improve decisions, not just outputs
- Operating model implications: whether the business is adopting AI as a tool, workflow layer, or strategic system
Most businesses skip most of these. That is why most AI adoption produces uneven results.
The businesses that get the most from AI are not always the ones that moved first. They are the ones that were clearest about what AI was meant to improve, where human judgement needed to stay in control, and how they would know if it was working.
That sounds less exciting than buying tools. Which is probably why it works.
Ready to assess your AI readiness?
If you are considering AI adoption and want a clear view of where to start, Humaine can help you assess the commercial, human, and operating model decisions that need to be made first. Book a strategy session to understand where AI can create advantage in your business, and where it could create risk if adopted too quickly.
Frequently asked questions
What should a business do before adopting AI? Before adopting AI, a business should define what AI is meant to improve: decision quality, commercial intelligence, team capability, brand distinctiveness, or operational speed. It should also identify which human capabilities must remain in control, which workflows AI will change, and how success will be measured beyond time saved. Starting with tools before clarifying those outcomes is the most common and most expensive mistake.
What is the first step in AI adoption? The first step is not choosing a tool. It is defining the business outcome AI is meant to improve and deciding where human judgement must remain in control. Once those are clear, the right workflow, governance model, and measurement approach tend to follow.
How should a business prepare for AI adoption? By reviewing strategy, workflows, data quality, governance, brand standards, and team capability before selecting a platform. AI works best when it is introduced into a clear operating model with defined review points and measurable outcomes. Without that foundation, it tends to create more activity without more advantage.
What questions should leaders ask before adopting AI? What is AI meant to improve? Which human capabilities must remain active? What decisions will AI support? Is the operating model ready? How will success be measured? Those five questions do not delay adoption. They prevent the kind of tool-led experimentation that produces cost without clarity.
Why do AI adoption projects fail? Most fail because they start with tools rather than outcomes. Businesses automate unclear workflows, measure activity instead of value, and underestimate the need for human judgement, governance, and change management. AI does not fix weak strategy or poor data. It tends to make both more visible.
How do you know if your business is ready for AI? When it has a clear outcome, a defined workflow, usable data, agreed governance, and a measurement model that goes beyond efficiency. It also needs to know where human judgement, creativity, empathy, taste, and meaning-making must remain active. If those decisions are still open, readiness work should come before scaling.
Should businesses start AI adoption with tools or strategy? Strategy first. Tools matter, but only after the business has defined what AI is meant to improve and how it will be governed. A tool-first approach tends to produce disconnected experiments, inconsistent quality, and ROI that is difficult to trace.
How can a business adopt AI without damaging brand quality? By keeping human judgement in control of strategy, angle, voice, and publishing decisions. AI can support execution, variation, and workflow efficiency, but it should not decide what the brand believes or how it sounds. That is where creativity, empathy, taste, and meaning-making need to stay human.
What should be measured after adopting AI? More than hours saved or content produced. Decision quality, speed to insight, brand distinctiveness, pipeline quality, team capability, and commercial impact are all worth tracking. Efficiency is a useful signal. It should not be the only one.
What is the biggest mistake companies make when adopting AI? Treating it as a technology rollout rather than an operating model decision. AI changes how work is created, reviewed, governed, and measured. When those decisions are not made deliberately, it is possible to increase output while quietly weakening quality, accountability, and strategic clarity.