The Next AI Wave May Not Write Anything
Key takeaways
- TypeSafe AI's Jev model and OpenAI's Decisions API represent a new category of AI built to select from predefined options rather than generate text.
- When AI decisions become cheap and near-instantaneous, automating judgement calls that were previously uneconomical becomes viable at scale.
- AI decision models may prove most valuable at process handoff points, where work waits for someone to decide what happens next and bottlenecks accumulate.
- Generative AI typically automates the outputs of work, while decision models may automate the transitions between steps.
Two product launches within two weeks suggest that AI is beginning to specialise in small, recurring decisions: the ones that move work from one step to the next. TypeSafe AI's Jev model and OpenAI's Decisions API both select from predefined options rather than generating text. The more interesting question is why this category is emerging now.
Three observations:
- Most business processes are built on small decisions. A customer message gets routed to the right team. An invoice passes through or goes to review. None of these require text generation, but every one of them requires judgement.
- Cost and speed change the scale of what's worth automating. When a decision is cheap and arrives in a fraction of a second, it becomes viable to automate judgement calls that previously weren't worth the effort.
- The decision becomes its own building block. What's new is that the decision is separated out as a distinct software component: here is the situation, here are the permitted options, choose one and report how confident you are.
In practice, this could look like the following in sales. A new lead arrives, and someone has to assess whether the prospect is ready to engage right now and which sales rep should take it. That assessment might sit waiting for days while interest cools. A decision model could make the call immediately, routing hot leads straight to a rep and flagging uncertain ones for human review.
I noticed the same pattern in a consulting project I recently completed. The most compelling AI use cases turned up at the handoff points, where a task moves from one person to another. Those same points were also where the bottlenecks sat. Generative AI typically automates the outputs of work. Decision models may turn out to automate the transitions between steps.
Where in your organisation does work sit waiting for someone to decide where it goes next? And which of those decisions would be the last one you'd hand to an AI?
#AIStrategy #EnterpriseAI #AIAgents #HavuAI
Marko Paananen
AI consultant and builder with 20+ years in digital business development. Helps companies turn AI potential into measurable business value.
Follow on LinkedIn →Related Insights
Distributed AI Development: Strategy or Drift?
Distributed AI development can be a sound strategy, but if no one maintains the overall picture, competitive advantages might not be identified.
The Second AI Investment: Systems, Not Models
Most organisations have made their first AI investment. The second, building reliable systems around models, is where the real work begins.
AI's value moved from models to workflows - where organisations should focus
The bottleneck in AI is no longer model capability but workflow judgment. Organisations must identify where repeated decisions slow the business down.
Interested in learning more?
Contact us to discuss how your company can leverage artificial intelligence.