The right order is the reverse of the common one. Not “buy AI”, but first put the processes in order: formalize and digitise them, build a methodology — and only then run AI agents inside the processes.
«Automation applied to an efficient operation magnifies the efficiency. Automation applied to an inefficient operation magnifies the inefficiency.»Bill Gates, “Business @ the Speed of Thought”, 1999
Make the process explicit: steps, roles, approval rules, documents and where the history is kept. An informal process lives in email and people’s heads — AI has nothing to stand on there. This is the foundation: structure first, intelligence second.
One source of truth, granular permissions and audit. Per RAND, common root causes of AI failure are fragmented, unready data infrastructure and a lack of governance.
«More than 80% of AI projects fail — twice the rate of non-AI IT projects; causes include unready data and missing governance.»RAND Corporation, 2024
Don’t bolt agents onto legacy — rebuild the process so AI works inside it. McKinsey found that redesigning workflows has the biggest effect on AI’s impact on profit.
«The redesign of workflows has the biggest effect on an organization’s ability to see EBIT impact from its use of gen AI.»McKinsey & Company
AI value is created not in the lab but on the front lines of the business. Per BCG’s “10-20-70”: 10% of the effort is the algorithm, 20% is data and technology, and 70% is people and processes. Technology is secondary; reshaping the work is primary.
«10% building the model, 20% high-quality data and technology, 70% new business processes and transforming how functions operate.»BCG, “Where’s the Value in AI?”
Pick one high-volume process with a clear owner and visible pain, put an agent inside it, prove the value with real metrics — and scale by function. That is the DATM methodology: from discovery to production in governed steps.
See the DATM methodology →