Why it matters now
The market is full of products that call themselves AI agents, and many internal teams are keen to build their own. Both can waste a year. The question I ask is simple: does this capability set us apart? If not, buy it. If it does, it may be worth building, but only where your data and processes give you an advantage nobody else has.
- 31% of AI technology budgets go to internal research and development.Wharton and GBK Collective, 2025 AI Adoption Report ↗
Common mistakes
- Building capabilities in-house that vendors already do well.
- Buying products labelled as agents that do little more than a chatbot.
- Signing up without an exit option or a plan for your data.
- Choosing on demos instead of on fit with your workflow.
What to do instead
- Buy the commodity. Writing assistance, meeting notes, search and similar tools are widely available. Buy them.
- Build where you differ. Build where your proprietary data or process is the advantage, and where off-the-shelf tools cannot reach it.
- Partner for skills. Use partners to borrow experience you cannot hire quickly, and make sure the knowledge stays with your team.
- Test on your own work. Trial every option on your real workflow and data before committing. Check that it learns and improves.
- Plan the exit. Know how you would switch vendors and get your data back before you sign.
Your first step this week
Take your current list of AI tools and planned builds, and mark each one “commodity” or “differentiating”. Question every build marked commodity.
Related guides
- Is our data ready for AI, and what foundations do we need?
- How fast should we move on AI, and how much should we invest?
- How do we get from AI pilots to scale?
All AI leadership principles and guides · 100 AI use cases by function