Managing Inertia in AI Adoption
Adopting artificial intelligence (AI) meets the same 16 forms of inertia as any other change. This page gives an AI example of each form and the counter that fits it. For what inertia is, where it comes from, and how to spot it on a map, read the main guide, Managing Inertia in Wardley Mapping, linked at the foot of this page.
Before You Start
Inertia is resistance to change, and it usually comes from past success: the more you have invested in how you work today, the harder that investment pushes back. Each form below comes with a counter, which is the tactic and the argument you prepare in advance so that you already have an answer when the objection comes. AI is not one component on a map. For many organisations, access to a general-purpose model through an API already behaves like a product or utility, while an assistant built around your own processes and data is still Custom-Built. Map the AI components you are changing and place them on the evolution axis first. That tells you which forms of inertia to expect.
Disruption of Past Norms
These forms are about what you already have and stand to lose.
Change of Business Relationship
Example: your long-standing IT services partner has little to offer in AI, and the people who manage that relationship stand to lose influence. Counter with vendor management: the partnership was right for its time, other teams have already brought in new suppliers, and the relationship owners can lead the change rather than be bypassed by it.
Loss of Existing Financial or Physical Capital
Example: you recently paid for licences or hardware for a rules engine, a search tool, or an analytics platform that a model could now replace. Counter with future planning: plan the write-down, run the old system down rather than extending it, and compare the cost of change with the rising cost of maintaining the legacy.
Loss of Political Capital
Example: the executive who sponsored a recent automation programme sees AI as a judgement on that decision. Counter with modernisation: present AI as building on what came before, focus on future agility and efficiency, and make the wider business aware of why the change is happening.
Threats to Barriers to Entry
Example: your advantage rested on specialist knowledge or expensive tooling, and AI lets smaller competitors do similar work for less. Counter by treating it as unavoidable change: show where it is already happening in your market. The barrier will fall whether or not you take part.
Transition to the New
These forms are about the cost of moving to the new way of working.
Investment in Knowledge Capital
Example: staff need to learn how to work with AI tools, how to check their output, and when not to use them. Counter with training: hiring these skills from outside will often cost more than developing your own people, and training shows staff that the change includes them.
Cost of Acquiring New Skill Sets
Example: nobody in the organisation can evaluate models, build retrieval over your own documents, or test an AI system for safety and bias. Counter with organisational development: grow the capability in-house through hack days, conferences, and a centre of gravity where the people learning these skills share what works.
Investment in New Business Relationships
Example: you need model providers, data partners, and specialist integrators you have never worked with. Counter with vendor management: spend time understanding the market and building relationships with the right suppliers before you need them.
Changes to Governance, Management, and Practices
Example: approval processes, risk controls, and data policies assume software gives the same answer every time, and generative AI does not. Counter with awareness of co-evolution: practices change as activities evolve, as DevOps did when computing infrastructure became a utility. Update governance alongside the technology, not after it.
Agency of the New
These forms are objections to the new thing itself.
Suitability
Example: "AI isn't accurate enough for our work" or "it isn't ready for regulated processes". Sometimes that is true; often it is inertia. Counter with weak signals and prior identification: examine the component's ubiquity and certainty. How widely is it used for this kind of task, and how well understood are its limits? If both are rising, the objection will not hold for long, so prepare now rather than wait.
Lack of Second Sourcing Options
Example: the system you want depends on one provider's model, and switching would mean rebuilding it. Counter with supply chain management: use standards and open source options where they exist, limit your use of features only one provider offers, and put an abstraction layer between your application and the model so you can swap it.
Lack of Pricing Competition
Example: the capability you need is available from only a few providers, so you fear being overcharged once you depend on it. Counter with market analysis: check how many vendors offer it, work out what switching would cost you, and consider whether a broker or reseller could get you better terms.
Loss of Strategic Control
Example: running a core process on an outside provider's model feels like handing over control of something that matters. Counter with strategic planning: examine the buyer and supplier relationship, including what happens if the provider changes its terms. Where the underlying capability is commoditising, keep control of what you build on top of it, such as your data, your processes, and your relationship with users.
Business Model
These forms come from how you make money, and from the people and markets that reward it.
Declining Unit Value
Example: you sell work such as translation, document review, or first-line support, and AI lowers the cost of each unit. Counter with awareness of evolution: defending the old price can become a death spiral as customers move to cheaper options. Look for alternative opportunities, such as new services built on the cheaper capability.
Data for Past Success Counteracts
Example: the figures that justified your current operating model make an AI alternative look risky, because they measure what worked before. Counter with portfolio management: treat AI work as a portfolio of small bets, fund them separately, judge them on their own measures rather than the legacy business's, and be ready to spin a successful one out.
Resistance from Rewards and Culture
Example: people are rewarded for headcount, billable hours, or personal expertise, and AI threatens all three. Counter through human resources: change what you reward so that adapting pays, explain why the change is happening, and promote situational awareness by sharing the map with the people affected.
External Financial Markets Reinforce Existing Models
Example: investors value you on your current margins and may react badly to AI spending before it pays back. Counter with analyst relationships: explain the long-term case with a clear story about where your market is heading. If you are a listed company, do this within your disclosure obligations.
Can AI Help Reduce Inertia?
A little, and only alongside the work above. AI can summarise feedback from the people affected by a change, draft training material for different roles, or analyse operational data to test the claims made for and against the change. It cannot remove inertia for you. Inertia comes from people's real interests, and dealing with it takes the conversations, pilots, and shared maps described in the main guide.
Sources
- Adapted from Simon Wardley, A pet favourite - Inertia, licensed under CC BY-SA 3.0.
- Adapted from Mark Craddock, Understanding and Managing Inertia in AI Implementations.
Material adapted from these sources is shared under CC BY-SA 4.0.