The race for the most sophisticated AI model misses the point entirely. Whoever owns the data will win in the agentic AI landscape, regardless of which foundation model sits underneath. Companies obsessing over GPT-5 versus Claude versus Gemini are fighting yesterday's battle whilst their competitors build competitive moats around data control.
Ula Nairne from Stelia cuts through the model hype with uncomfortable clarity in our recent conversation on The Precision Brief. Most marketing teams are paying three times over for the same AI capabilities because they focus on technology sophistication instead of data architecture. The winners understand that sustainable AI advantage comes from proprietary data assets, not rented model intelligence.
Individual prompting is dead — enterprise needs orchestration
The ChatGPT era taught marketers the wrong lessons. Individual prompting worked when AI was a personal productivity tool. Enterprise AI requires orchestration across systems, data sources and decision chains. Most companies are still thinking in terms of clever prompts when they should be building data pipelines.
Token costs are now marketing unit economics. Every API call, every data transformation, every model query shows up in campaign ROI calculations. CMOs who treated AI as a free lunch discovery are learning that scale brings cost structure decisions. The bill arrives monthly and it scales with usage, not results.
This shift forces marketing leaders into infrastructure conversations they never expected to have. Token budgets. Model switching costs. Data residency requirements. The CMO who cannot discuss these topics with confidence will lose budget control to whoever can.
Finance teams are applying traditional procurement logic to AI spend. Predictable costs beat impressive demos every time. Companies that built AI strategies around experimentation budgets are hitting CFO resistance when those experiments need to scale into line items.
Agency AI platforms create particularly expensive black boxes. Impressive interfaces that hide model switching, token consumption and data processing costs. Marketing teams get beautiful dashboards whilst finance teams get unpredictable invoices. The sophistication becomes a liability when nobody can explain what drives the monthly bill.
Most companies waste money on advanced models for basic tasks. GPT-4 for simple classification. Claude for data extraction that rules-based systems handle faster and cheaper. The model sophistication arms race distracts from workflow optimisation that actually moves performance metrics.
The companies that will dominate agentic AI already understand this shift. They optimize existing workflows instead of building new systems from scratch. They control their training data, their inference data and their performance feedback loops. Model choice becomes a technical detail, not a strategic decision.
Enterprise AI has two core problems that sophistication cannot solve: governance and cost control. Governance means knowing what data feeds which decisions at what cost with what approval chains. Cost control means predictable unit economics that scale with business value, not computational complexity.
Agencies selling AI transformation programmes miss this entirely. They lead with model capabilities instead of data ownership strategies. They build dependencies on their platforms instead of transferable capabilities inside client organisations. The sophistication impresses procurement committees but creates vendor lock-in that finance teams recognise too late.
Watch the full conversation on The Precision Brief to hear Ula's complete framework for building sustainable AI advantage through data control rather than model sophistication. Subscribe to our newsletter for weekly insights on what actually moves the needle in data-driven marketing.