Blog: Technology choice
The best AI model is a business decision, not a league table
Published by Forever Foundry
AI buyers are increasingly being asked to choose a winning model before they have defined the work it needs to do.
The benchmark is impressive. The demonstration looks convincing. The licence is available.
But is it the right approach for the business problem?
Different tasks may need different AI
On 22 September, Palo Alto Networks announced a new security service that uses a “multi-model harness”. Rather than applying one AI model to every task, its architecture can route work between Anthropic’s Claude Mythos 5, OpenAI’s GPT-5.6-Cyber and open-weight models.
Palo Alto Networks says this is intended to improve security coverage while controlling the cost of using frontier AI at scale. It combines those models with threat intelligence, testing methods and human expertise. Read the announcement.
These are claims made by the vendor about a specialised cybersecurity service, not an independent comparison or evidence that a multi-model approach is right for every organisation.
The FOFO™ view is that the important idea sits beneath the product announcement: the work should determine the technology, not the other way round.
One organisation can contain very different requirements
Imagine a manufacturer considering AI across three areas.
The quality team wants to identify recurring faults. Procurement wants earlier warning of supplier problems. Customer service wants to respond to routine enquiries more quickly.
All three could be described as AI opportunities. They are not the same problem.
Quality may require high accuracy, explainable evidence and human approval. Procurement may depend on combining internal and external information. Customer service may value speed but need strict limits on what can be promised.
Selecting one model or platform first can force those different problems through the same technical answer.
Listening across the organisation changes the starting question from:
“Which AI should we adopt?”
to:
“Which business problem deserves investment, and what would a successful solution need to do?”
The answer may involve one model, several models, existing software, process redesign or no AI at all.
Define the outcome before comparing the options
Before selecting technology, give each opportunity a clear decision basis:
- What operational outcome should improve?
- What errors would be unacceptable?
- What information and permissions would be required?
- How quickly must the task be completed?
- What would it cost per successfully completed case?
- Who owns the outcome and how will improvement be measured?
Those questions turn model selection into a delivery decision rather than a strategy.
The FOFO Process™ begins earlier. AI Compass™ listens across the organisation to uncover the challenges and opportunities people can see. Slid3rs™ keeps prioritisation human, helping teams rank them by business value and delivery complexity. FlowRoom™ brings leadership together to agree what should happen first.
Only then should the organisation compare delivery options.
The same principle applies when the next step is a supplier brief. Read why AI-written requirements still need the right problem behind them.
There may be no single best AI model for your business. There are business problems, required outcomes and constraints, and technologies that fit them to different degrees.
Choose the problem first. Let the evidence determine what comes next.
Next