A Framework for AI Adoption — And Why Most Budgets Are Funding Hobby Projects Instead

The real barrier isn’t the technology, it’s the CxO’s dilemma of figuring out what problem it solves and what the ROI is. Most enterprises are still stuck on that question, and the budgets that do get allocated too often fund an executive’s pet project instead of a scalable capability. That’s the gap this framework is meant to close.

Start with two reframing questions rather than a vendor RFP – if you had extra headcount budget, where would you deploy it – that tells you where the real bottleneck is. And stop treating AI as a compute problem the way you would traditional software; treat it as a prediction problem and ask what you’d like your data to predict for you.

Here is the AI Adoption Maturity Framework I use with clients, in four levels.

Level 0, ready now: Pointed AI solutions solving one specific, well-proven problem with well established techniques and technologies – NLP, computer vision, LLMs applied to data extraction and cleanup across ERP and home-grown systems – that slot into existing architecture with no major redesign.

Level 1, ready now and evolving fast: A cognitive layer overlaid on your architecture that introduces predictions inside existing workflows, for example predicting sales at the moment inventory is received while a separate model drafts the marketing plan around it.

Level 2, ready in 6 to 12 months: AI agents as team members, not co-pilots that are added to a product team alongside the PM, developers, and tester to accelerate delivery, though this only works once Level 1 is genuinely in place.

Level 3, ready in 18-plus months, possibly sooner: AI-first delivery, where an AI-agent team executes the project and humans are added for oversight, governance, and quality assurance.

The mistake I see most often is enterprises trying to skip straight to Level 2 or 3 for the press release, without Level 1 in place, which is exactly how budget turns into a hobby project. Map your current initiatives to these four levels before your next planning cycle; if most of them cluster at Level 0 while your public messaging implies Level 2, that gap is worth closing quietly before a board member notices it.

The bigger disruption from this framework won’t be inside the enterprise. It will hit the engagement models of IT services providers like Accenture, Infosys, TCS, and others. That’s the subject of a later post.