It’s most likely of no surprise to you that marketing practitioners are leading the way in adoption of AI, be it at the agency level or within the enterprise. Extensive research from Salesforce, Mckinsey, Hubspot, and ZoomInfo, among others peg markets between 65% to nearly 90% at adopting some form of AI in their day-to-day workflow.
It makes sense – AI can be very good at high volume, repetitive tasks. There is also a very low technical barrier – most marketing tools ae easy to deploy and use, and the outputs, things like writing and design, require zero engineering integration like code.
And let’s get real here about us marketers – we are always looking for an edge and willing to try new things. It’s the nature of who we are.
What you might not know is that marketing, in its race to adopt to be more efficient and competitive, is also leading in creating significant risk in their organizations. The numbers making this case are staggering:
- The IAB reports that 70% of marketers have already encountered an “AI Incident” that includes hallucination, off-brand content, or bias.
- Almost 60% use unsanctioned tools (known as ShadowAI) at work. Almost 70% do work on personal accounts (off their company network) and almost 60% put sensitive company data into public LLMs. Only 23% use the governed, company sanctioned tools. To be clear – this is bad. Very bad. Everything from IP, to PII, to proprietary data, to accurate outputs, is at risk.
- Shockingly, only 35% of those cited in the IAB study plan to invest in oversight in the next 12 months in spite of incidents that put their organization at risk. Hat tip to all my marketing friends out there — willful ignorance risk is not a defense for when the inevitable AI created crisis happens.
- More than 80% of marketers are actively piloting agents – most without the necessary oversight. Per the Harvard Business Review “poorly governed agents amplify errors and can generate coordination failures that break adjacent workflows.”
What’s crazy about all this – the solution is simple and cost effective. Oversight and governance tools are out there, and even the most basic ones can eliminate a good percentage unsanctioned (ShadowAI) use, or restrict the release of PII and IP. My strong, unsolicited advice to you marketing leaders out there – this is not optional.
So, in the spirit of “fix the problem, not the blame,” here’s some thoughts about what you should be looking for to provide oversight of your AI consumption – for both employees and agents:
- Look for an “AI-Native” tool – meaning, it was built specifically for AI oversight. Legacy security providers don’t yet have the chops to effectively oversee AI.
- Platform maturity is important. New companies are popping up every day; few have a track record of success. Since this is a relatively new category, looking for oversight companies that have had active, happy customers for at least 2 years or more.
- Your Data Should be Encrypted and Secure – Most of the AI security tools will analyze and store AI traffic. Make sure that any storage is encrypted and secured at the highest standards to ensure your data stays protected. (NIST and FIPS 140-2 are suggested)
- In line controls are essential – Your oversight should enable real time prevention in allowing risky prompts from both employees and agents.
- Real-time, Zero-Day ShadowAI & Shadow Agent detection is a must – New, unsanctioned tools, come out every day – you can catch them before they become an unmanaged tool in your organization with zero-day shadow AI and Agent detection. Most oversight companies catalogue these sites monthly and don’t catch all of them, so be sure to look for real-time detection.
- Cost/Token Controls are critical – With autonomous agents, controlling the costs of your AI deployments (number of Tokens used) has now become an imperative. Recursive loops (agents going back and forth with each other to solve problems), mission drift, and too much agency are just a few things that are proven to escalate costs and rapidly exhaust budgets.
- There’s Power in the Data/Analysis of your AI Consumption — The data used for detecting risk is also extremely valuable in discovering innovation and opportunities. Look for an oversight solution that provides both defense to protect AND offense to take better take advantage of AI and achieve the desired ROI. Best in class companies will enable you to understand use cases, what agents to build and measure KPIs – even cost effectively manage your licenses.
While you and your organization might have additional requirements, the above 7-point rubric are the bare minimum for effective AI consumption oversight for marketing departments and agencies. As marketers, we should not just take pride in how we lead in adoption, we should also lead in the oversight and management of these important new tools.
