The Real State Of AI May 2026
On the AI Moment podcast, we wanted to tackle the real state of AI from the end of April into early May. Here are the 10 most important themes we covered and the action to take.
I have embedded the podcasts at the bottom of the post to go deeper and hear the other themes we discussed.
1. Set a Clear Vision for the Future of Work
Explainer: The mass of the workforce currently fears that AI will take their jobs, which stifles innovation, and many executives lack a definitive "future of work" story. Without a guiding vision, businesses see disjointed, cost-saving experiments rather than strategic transformation.
Action to take: Senior leadership must define and communicate a clear 3-year vision detailing how AI will be integrated into the organisation to augment staff and achieve new capabilities, giving employees a "shining castle on the hill" to work towards.
2. Map Workflows Before Automating
Explainer: Applying AI tools to poorly understood or informal processes will simply speed up existing inefficiencies. AI agents moving at high speeds will quickly expose weak governance and messy internal workarounds.
Action to take: Document your exact workflows and clean up operational inconsistencies department by department before deploying AI. Understand exactly what your teams do and how they do it so you are automating a refined process, not a broken one.
3. Rethink Budgeting for Token Consumption
Explainer: Traditional, static annual budgets clash with AI consumption models. Companies using AI aggressively are rapidly burning through token allowances, with examples like Uber exhausting a year's budget in just four months.
Action to take: Transition to agile financial planning for AI. Create flexible budgets or an "emergency pot" of funds that can be unlocked dynamically to scale successful AI deployments without derailing the company's financial planning.
4. Establish Robust Governance and Sandboxing for AI Agents
Explainer: AI agents can execute complex tasks in seconds, but if deployed into environments with weak access controls, they become major security liabilities vulnerable to prompt injection or catastrophic errors.
Action to take: Treat AI deployment like a driving test. Run AI agents in isolated sandboxes to test their behaviour and ensure strict governance validation before granting them access to live company networks.
5. Redesign Roles to Augment Humans, Not Just Cut Headcount
Explainer: Freezing hiring with the vague hope that AI will replace staff is a major misstep. AI replaces specific tasks, not entire jobs, and misunderstanding what employees actually achieve leads to operational failure. Middle management is rightly concerned, however, once AI systems struggle to understand and translate what leadership wants and how to connect it with more junior or more technical staff, this will realign.
Action to take: Focus on clever role redesign that layers AI into workflows to handle repetitive tasks, freeing your human workforce to handle higher-order thinking and tasks requiring "gut feel" and human judgment.
6. Protect the Development of Junior Talent
Explainer: Handing over all low-level analytical work to AI risks creating "cognitive debt." If junior staff do not practice foundational skills, the business will eventually lack capable senior leaders who understand the nuances of the industry.
Action to take: Use AI to accelerate junior learning, not replace it. Ensure junior employees still take ownership of final outputs and use AI to augment their research, allowing them to learn necessary patterns faster while remaining under senior supervision.
7. Shift from "Productivity Theatre" to Measurable ROI
Explainer: Many businesses are generating noise and building code without seeing a real Return on Investment (ROI) because deployments aren't tied to strategic goals.
Action to take: Stop running disconnected 6-to-8-week deployments. Implement short, 2-week sprint-based workflows to test AI solutions, measure their direct business impact, and only scale the token usage for those that demonstrate undeniable success.
8. Assume Zero Data Privacy in Public Models
Explainer: AI platforms are increasingly using user inputs to train their models, and the boundary between private corporate work and public training data is rapidly eroding.
Action to take: Operate under the assumption that anything done on public online networks is not private. Implement firm internal data policies, so employees know what proprietary company data can and cannot be used with external AI tools.
9. Build Your Company's Internal "Wiki"
Explainer: You cannot build successful automated agents if your company lacks operational documentation. Wiki’s lost it’s importance to company Slack/Teams and internal search, but wikis are incredibly important, whether you build in Notion, Confluence, or host something similar on Google drive etc. AI tools require a deep, structured understanding of your business to function properly.
Action to take: Invest time in becoming operationally excellent by documenting your business. Create the foundational repositories and technical files needed so AI models can actually understand your internal context.
10. Upgrade from Software Licenses to Practical Hackathons
Explainer: Simply buying software licenses (like Copilot) and handing them to employees results in zero progress. Staff often understand the basics but don't know where to start applying AI to their daily jobs.
Action to take: Deploy guided workshops and "hackathons" that force teams to apply AI directly to a real-life business problem. Once employees solve one tangible problem with AI, adoption accelerates naturally.
