Blog: Supplier dependency
The fastest AI build can create the longest dependency
Published by Forever Foundry
A supplier can build an impressive AI agent remarkably quickly.
The harder question is what happens six months later, when the process changes, the original engineers have left and nobody inside the organisation knows how to adapt it.
Speed to deployment matters.
So does the ability to take ownership afterwards.
Early progress can hide long-term dependency
On 30 September, Gartner predicted that 70% of enterprises will abandon agentic AI created through vendor “forward-deployed engineering” by 2028.
This delivery model places vendor engineers directly alongside the customer to build and integrate a solution. Gartner argues that some organisations will face rising costs and find themselves unable to evolve what has been built without continued supplier involvement.
It recommends agreeing governance, business-value expectations, intellectual-property ownership, knowledge transfer and an exit plan before signing. Read the coverage of Gartner’s prediction.
This is a Gartner forecast, not an observed failure rate or a prediction about every supplier-led AI project.
The FOFO™ view is that technical speed becomes dangerous when it gets ahead of organisational understanding.
A working agent can still solve an incomplete problem
Imagine a manufacturer commissioning an AI agent to manage quality exceptions.
A specialist vendor may be able to create a convincing prototype from the quality team’s requirements. But the complete outcome crosses several functions.
Operators know where information first goes missing. Quality understands which exceptions require judgement. Operations sees the effect of waiting. IT understands the systems and dependencies. Customer service experiences the consequences when approval delays affect an order.
If those perspectives are missing, the agent can automate the version of the problem described by its sponsor without addressing the problem experienced across the business.
That creates two forms of dependency.
The technical dependency is needing the supplier whenever the solution must change.
The more serious dependency is relying on the supplier’s original interpretation of how the organisation works.
The same applies once an agent takes on ongoing work. Read about AI agents being given ongoing responsibilities.
Design for ownership before delivery begins
Before appointing a supplier, establish:
- The complete business problem and intended outcome
- The people and functions affected by the change
- The operational baseline and measures of success
- Which decisions remain with people
- Who owns the process, data and resulting capability
- What knowledge must be transferred
- How the organisation will change or retire the solution
Measure more than delivery speed.
Track time to change the solution, exceptions requiring manual intervention, cost per successful outcome and the amount of external support needed after launch. A fast implementation is not necessarily a low-cost one if every adjustment requires another supplier engagement.
The FOFO Process™ establishes the organisational evidence before delivery options are chosen.
AI Compass™ listens across the organisation to uncover the complete challenge. Slid3rs™ helps the people affected rank the opportunity against competing investments. FlowRoom™ brings leadership together to agree the outcome, ownership and roadmap.
Only then should the organisation decide who should deliver it and how.
The right supplier can accelerate progress enormously. But the organisation should finish stronger, not simply more dependent, because that supplier was involved.
Finance should be part of that choice from the start. Read why your CFO should help choose the AI problem.
Discover and prioritise the business problem before choosing how to deliver it.
Related reading
Next