Step 3: Develop a Vision and Strategy

A vision for ethical AI is not a policy document. It is an answer to the question:  what kind of organisation do we want to be in relation to AI?

Kotter’s third step asks leaders to develop a clear, compelling vision of the desired future: one that is simple enough to be communicated in five minutes, ambitious enough to inspire, and concrete enough to guide decisions.

For ethical AI, a powerful vision might look like:  “We build and deploy AI that we would be proud to explain to anyone it affects.” Or:  “Our AI systems are transparent, accountable, and designed with the dignity of every person in mind.”

Vision is not the same as a list of principles. Many organisations have published AI ethics principles that have had no discernible impact on their actual AI practices. Vision is what those principles look like when they are lived: in decisions, in trade-offs, in the moments when ethics and expedience conflict.

Strategy translates vision into direction. It answers:  how will we get from where we are to where we want to be? For ethical AI, strategy involves choices about sequencing, resourcing, governance structures, and measurement. It requires honest assessment of current capabilities and honest acknowledgement of current gaps.

Step 4: Communicate the Vision for Buy-In

Kotter identified under-communication as one of the most common causes of change failure. Leaders announce a new initiative, communicate it once or twice, then move on: while the organisation, unsure whether this is real or just another management initiative, waits it out.

Communicating a vision for ethical AI requires consistency, repetition, and: crucially: leadership behaviour that matches the words. If a leader speaks about ethical AI in the all-hands meeting and then approves a rushed deployment that skips the impact assessment, the message received is that ethics is optional.

Effective communication of an ethical AI vision uses multiple channels, involves multiple voices, and tells stories. Not just “we are committed to ethical AI” but “here is what that commitment looked like when we slowed down a deployment because we weren’t satisfied with our bias testing.” Specificity builds credibility.

It also involves creating two-way conversations. People need the opportunity to ask questions, raise concerns, and contribute to the vision: not just receive it. Organisations that communicate ethical AI as a mandate tend to generate compliance. Organisations that communicate it as a shared purpose tend to generate commitment.

Reflection question:  If you had to articulate your organisation’s vision for ethical AI in one sentence, what would it be?