Enterprise AI strategy needs more than ambitious experiments: it needs a clear business purpose, workable processes and people who can manage risk before and after launch. In a TechRadar Pro opinion article, Paul Maker, Aiimi’s chief technology officer, borrows a phrase from House of the Dragon to describe IT, Legal and Compliance as “protectors of the realm.” The fantasy metaphor is useful when it points to shared responsibility—not when it suggests those teams can guarantee safe outcomes on their own.
Why an AI strategy needs protectors of the realm
Maker’s argument is that IT, Legal and Compliance should help turn an AI ambition into an operating capability. IT can assess how a system connects to existing technology and workflows; Legal and Compliance can surface obligations and risks that shape how it should be used. Their involvement is most useful when it begins during design, rather than after a system is already embedded in a process.
![House of the Dragon: The Complete First Season (4K Ultra HD + Blu-ray) [4K UHD]](https://m.media-amazon.com/images/I/51JeAIckcXL._SL160_.jpg)
![House of the Dragon: Season 1 [Blu-Ray] [2022] [Region Free]](https://m.media-amazon.com/images/I/51Od+cjqc2L._SL160_.jpg)
As an Amazon Associate I earn from qualifying purchases.
That is a case for coordination, not a veto by default. Governance is one part of a wider strategy, and the right balance between speed and control depends on the workflow. A team handling low-consequence internal drafts faces a different cost of error than one using AI to inform decisions with legal, financial or personal consequences. Compare the cost of a mistake with the cost of delaying the use case before deciding what review, testing and human oversight it needs.
Free tools Windows power users keep installed
One-click scans. No signup required.
Start with the business outcome, not the model
Microsoft defines an AI strategy as “a coordinated plan for using AI to reach specific organizational goals.” Its guide organizes that plan around five connected threads: vision, data, people, processes and governance. Deloitte similarly advises tying AI work to corporate strategy and key performance indicators and involving leaders across divisions. These are recommendations from the respective organizations, not a guarantee that any particular AI initiative will succeed.
#1 Best Overall
Translate the business goal into a problem a team can observe and measure. For example, an organization might want to reduce the time employees spend applying data labels, improve the consistency of classification, or make it easier to find information under existing access rules. Before choosing a model or platform, determine how the current process works, who depends on it, what data it touches and what a meaningful improvement would look like.
- Strategic fit: Which business priority does the use case advance?
- Expected outcome: What measure would show that the workflow improved?
- Cost of error: What happens if the output is wrong, incomplete or misused?
- Cost of delay: What is lost if the organization waits, and is that loss greater than the risk of moving too soon?
Choose feasible use cases and build the foundations
A long list of AI ideas is not a strategy. Prioritize opportunities that connect to business goals and can be supported by the organization’s data, skills, systems and risk controls. A promising use case may still need better data access, workflow changes, staff training or a clear escalation route before it is ready to scale.
Check the data and obligations
Identify what information the system will use, who is permitted to access it, how sensitive it is, and what legal or compliance requirements apply to the intended use. Establish how outputs and inputs are handled, who can review them, and what records the organization needs to keep. Legal and Compliance can help interpret obligations for the specific context; a general AI policy is not a substitute for that assessment.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsFit AI into a real workflow
Map the steps around the AI task: what a person supplies, what the system returns, who checks the result, and what happens when it fails or cannot answer. Integration matters because an isolated tool may create extra work or bypass established controls rather than improve the process. IT should help assess the technical connections and operational dependencies before launch.
Assign people, skills and ownership
Name a business owner accountable for the outcome and operational owners responsible for the system, data, oversight and escalation. Identify who needs training, what decisions remain with people, and how users can report unexpected behavior. OpenAI’s business guide describes organizational adoption through five principles: communication, role-specific training, experimentation, scaling and acceleration, alongside governance. Its principles are a useful adoption framework, not evidence that one sequence fits every organization.
Use governance to shape the design, not just approve the launch
Governance is more practical when it answers concrete questions about a particular use case. The people responsible for technology, legal obligations, compliance and the business process can work together to define acceptable use, testing, access, human review and monitoring. The right controls will vary with data sensitivity, likely harms, workflow integration and the consequences of an incorrect output.
Maker’s article proposes AI-supported data classification and governance as an example: design the use case with compliance and legal understanding, and connect it to IT-managed workflows. Treat that as a proposal for how cross-functional planning might work, not as a validated product recommendation or proof that the approach will deliver a particular result.
Measure results and keep oversight after deployment
Define success before a pilot begins. Choose measures tied to the original business goal, such as time spent on a task, consistency of results or the rate at which people must correct outputs. Record a baseline so that a change can be compared with the existing process, and consider whether the measure captures quality and risk as well as speed.
After launch, monitor whether the system still fits the workflow and whether users are following the intended process. Set out who reviews performance, who can change or pause the use, and how incidents or policy exceptions are escalated. A launch approval is not a substitute for ongoing ownership: data, systems, user needs and obligations can change.
Rank #4
Decide whether to build, buy or partner
The choice depends on the use case and the organization’s capabilities. Microsoft’s guide frames build, buy and partner decisions around tradeoffs that include control, speed, cost and differentiation. A current, use-case-specific evaluation is still necessary; no general strategy guide can establish which option is best for a particular organization.
- Build when the required capability and differentiation justify the added work and responsibility.
- Buy when an existing offering fits the workflow and its data handling, integration and controls meet requirements.
- Partner when outside expertise or capability is useful, while making ownership, access and accountability explicit.
For any route, assess strategic fit, expected outcomes, data sensitivity and access, workflow integration, oversight responsibilities, and the cost of error versus delay. The decision is not simply whether a model is technically capable; it is whether the whole operating arrangement can support the intended use.
Recommended Free Tools
Ask questions that lead to action
Executives often ask when generative AI will match their best employees, whether it is accurate enough, whether the CIO is moving fast enough, and what rivals are doing. Harvard Business Review authors Bharat N. Anand and Andy Wu suggest reframing those questions around how an organization can use generative AI effectively now despite its limitations, and how it can create competitive advantage.
Best Value
That shift keeps attention on choices leaders can make: which workflow is worth improving, what level of error is acceptable, what capabilities are missing, and what evidence would justify expansion. OpenAI’s December 2025 business guide reports figures of 5.6 times growth since 2022 in frontier-scale model releases, 280 times lower cost to run GPT-3.5-class models over 18 months, and adoption four times faster than desktop internet. Those are figures reported by OpenAI; the guide’s presented material does not provide enough methodological detail to treat them as independently established measures. They may describe a fast-moving field, but they do not determine whether a specific enterprise use case is ready.
Further reading
For a broader discussion of strategy and leadership in data- and algorithm-driven businesses, Harvard Business School describes Competing in the Age of AI: Strategy and Leadership When Algorithms and Networks Run the World, by Marco Iansiti and Karim Lakhani, as covering data, algorithms, digital networks, operating paradigms and competitive strategy. Harvard Business School AI Institute: Competing in an AI-driven world.
Sources: Paul Maker, TechRadar Pro, 25 August 2026; Microsoft, AI Strategy for Business Transformation; Deloitte US, How to Create an Effective AI Strategy; OpenAI, Staying ahead in the age of AI, 16 December 2025; Bharat N. Anand and Andy Wu, Harvard Business Review, November–December 2025.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
