CMOs: Don't lock your brand into agency AI platforms
Gartner expects 50% of advertising agencies' proprietary AI platforms to wind down or be obsolete by 2029. The driver: open-source AI stacks from hyperscalers that are cheaper, more customizable, and built to run across the enterprise - not just marketing.
By 2028, open-source platforms will support more than 75% of enterprise AI deployments, according to Gartner. That puts any brand locked into an agency-owned platform at risk on cost, flexibility, and scale.
Why this matters
"I don't hear [agencies] talking about those platforms … being an enterprise-wide AI platform," said Jay Wilson, vice president, analyst at Gartner for Marketers. "That's the big risk to agencies right now. They're still thinking at the marketing level."
At most companies, the CIO will own the enterprise AI strategy. If your AI foundation is tied to an agency's stack, you'll likely be cut out of the decision and stuck defending point solutions that don't fit the company standard. As Wilson put it: "A CMO who's embedding with a [holding company platform] is likely going to get disintermediated from that discussion."
The core risk for CMOs
Agency relationships churn. Your AI platform shouldn't. Migrating content, data, prompts, workflows, and integrations between closed platforms is slow, costly, and distracting. The bigger your AI footprint gets, the harder the exit.
Meanwhile, hyperscalers and open-source ecosystems keep improving. The value tilts toward interoperability, not walled gardens.
What to do now
- Avoid long-term lock-ins: No multi-year AI platform contracts. Start with proofs of concept. Keep termination-for-convenience rights without penalties.
- Demand portability: Ensure you can export data, fine-tuning artifacts, prompts, chains/agents, and evaluations in open formats. Require documented migration paths.
- Prioritize interoperability: Require open APIs, standard connectors, and support for major hyperscalers (e.g., model gateways, vector DBs, data lakes).
- Align with the enterprise stack: Involve the CIO, CISO, data leaders, and procurement early. Lock down security, governance, and compliance from day one.
- Own the strategy, rent the speed: Use agencies for acceleration, upskilling, and external perspective - not as the system of record.
- Protect IP and data: Clarify ownership of prompts, fine-tuned models, training data lineage, and evaluation datasets in your MSA/SOW.
- Measure outcomes, not demo magic: Define clear KPIs (cycle time, lift, CAC, LTV, A/B deltas), and require model evaluation and bias reporting.
How agencies can still win
Agencies remain valuable for speed to market, cross-client pattern recognition, and the outside-in view of the business. "What an agency provides that is not tech-dependent … is that outside-in perspective," said Wilson.
To stay relevant, agencies should focus on human strengths: creativity, innovation, and customer insight. Package small, composable accelerators instead of monolithic platforms, build on open standards, and guarantee portability. Keep human-in-the-loop quality control and performance analytics at the center.
This also means investing in talent. Cuts today risk weakening the very edge clients pay for tomorrow.
A quick due-diligence checklist for AI platform decisions
- Exit plan and data/model portability (including prompts, embeddings, and evaluation sets)
- Clear IP ownership and usage rights in contracts
- Total cost of ownership vs. hyperscaler/open-source alternatives
- Security, privacy, and compliance posture (PII handling, auditability)
- Model lineage, safety guardrails, and bias/quality metrics
- Performance SLAs (latency, uptime) and support
- Integration with enterprise identity, data platforms, and observability
- Pilot first: 90-day PoC with measurable business outcomes
Bottom line
Build your AI strategy on open, enterprise-grade foundations. Leverage agencies for acceleration and creative impact - not as your core platform. Keep your options open, your data portable, and your contracts flexible.
If you're formalizing your approach, start here: AI Learning Path for CMOs
For broader market context, see Gartner.
Your membership also unlocks:
