How Dr. Martens, Target, Sawyer, Natpat Combat AI Tool Sprawl
As enthusiasm for AI mounts, it has become inevitable that employees seek out tools to increase their individual productivity — often, without IT’s knowledge, creating shadow AI. The proliferation of tools sometimes means teams have free reign to buy or use what they want without governance or oversight.
This is creating massive tool sprawl inside consumer goods and retail enterprises due to duplicative programs and subscriptions. This not only causes unnecessary spending, but AI tool sprawl also makes risk management more difficult while introducing unauthorized data paths and expanding attack surfaces.
“The issue [of AI tool sprawl] is coming from too many overlapping experiments without [a] value thesis behind how to bring them together,’’ says Holger Kömm, a partner at Bain & Co.
CPG companies are especially structurally prone to sprawl because they might have marketing buying creative tools, sales buying retail media tools, and supply chain teams buying tools for forecasting actions and events. It goes all along the value chain of consumer product creation, Kömm says.
Also: Unilever, Newell, Levi's lead the CPG pivot toward agentic workforces
This results in fractured data and an inconsistent brand voice, he adds. “[It's] the single outcome you don't want to see, and this is what you get when you go uncoordinated in all the single solutions.”
Getting Ahead of Tool Sprawl
With operational AI initiatives expected to continue growing, many CGs and retailers say they are increasingly taking steps to control AI tool sprawl.
Dr. Martens, for one, has upgraded its tech stack. More than 70% of its platforms now provide native AI capabilities. The British footwear and apparel company sought guidance from its tech partner, OSF Digital, to help establish a layered hierarchy across its tech landscape to manage AI tools effectively.
“The bigger challenge is found at the data layer in terms of traceability, consistency and context boundaries to ensure reliable response from the tools,’’ says Graham Calder, chief technology and AI officer at Dr. Martens.
The company strategy is to leverage AI wherever people see an opportunity, and to do so within a human-centric framework while recognizing and managing the risks, Calder says.
OSF helped the company map out use cases, and from there, Dr. Martens developed a governance framework spanning the full technology lifecycle.
“For example, our vendor onboarding was upgraded to include an assessment of AI capabilities and risks — the policy being to get it right the first time rather than remediating after the event,’’ Calder explains. “We set up regular cross-company forums for sharing where people were finding value in AI," recognizing their experimentation, curiosity and innovation in delivering business benefits while protecting the brand against adverse impact.
As with any technology, shadow IT is something that every organization needs to recognize and manage appropriately in terms of risk and opportunity, Calder notes.
Retail giant Target has a number of AI tools in place because accelerating the use of technology, including AI, is a strategic priority in the company's next chapter of growth, says Devon Cox, VP of technology architecture.
“Over the past year, we've become more intentional about connecting that innovation through a common architecture and governance model,” Cox says.
Enhancing the customer, brand and employee experience has been the main priority.
“The point isn't to maximize the number of tools,’’ he adds. “It's to build the foundation to develop, deploy and operate AI responsibly at scale.”
To engage with merchants, Target offers AI-enabled forecasting and planning tools to help identify opportunities sooner, make faster and more confident decisions, and bring compelling products to shelves more quickly, Cox says.
At the same time, IT is moving away from isolated, one-off tools toward shared capabilities and reusable patterns that can work across the company so it can scale technology in ways that support future growth, Cox says.
AI tool sprawl can be a natural challenge when organizations move quickly, he notes. “That’s why we’re focused on connecting innovation through a common architecture and governance model, so the work can scale with the speed, consistency and control required for enterprise growth.”
Without governance, every team builds agents its own way, Cox points out. “With governance, every agent declares five dimensions up front: what role it plays, what triggers it to act, how much autonomy it has, what systems and data it can access, and how it will be monitored and corrected.”
IT supports teams with shared platform capabilities for registration, observability, evaluation, guardrails and cost management — rather than asking each team to reinvent them, he says.
“That common foundation helps Target move from individual AI experiments to reusable capabilities that can support growth across the business. It also helps us reduce duplication, optimize resources and make disciplined investment decisions as adoption grows.”
Limiting Access
Water filtration and first-aid products company Sawyer uses four external AI tools and a custom chatbot named Tom that was designed to help customers find answers to questions on its website before reaching out to customer service.
Even with a small number of tools in use, shadow AI is a definite concern, says Nick Loftin, IT manager.
“The growth rate of AI usage and the faster adoption leaves the window open for vulnerability,’’ he says. The company has hosted a few roundtable discussions about AI best practices, and employees have been told to use paid AI models for anything company related to avoid or limit risks of data either leaking or being used in training.
“Where we might deal with [AI tool] sprawl is if we start having lots of agents running tasks,’’ Loftin says.
IT limits who needs access to tools aside from Google Gemini, he says. “We also try to link any data that’s used for training or knowledge base to the same files across AI tools.”
For example, IT tries to avoid having the same files in separate repositories for its chatbot and phone system. “Otherwise, it becomes difficult to ensure they are providing customers with the same information,’’ Loftin says.
IT also does not allow blanket connections to shared folders. Whenever a new AI tool is explored, IT provides a subset of users access to see if it’s beneficial. If it is, then executives look at whether employees will still need access to an existing tool.
Ensuring Approved Tools Are Useful
Michael Jankie is mindful of the fact that employees like to bring in their own AI tools.
“In a fast-moving company, people will naturally use tools that help them get work done,’’ says the co-founder & CEO of Australian-based Natpat, which manufactures natural wellness patches and bug protection. “The risk is not that people are curious; the risk is when sensitive data, customer information, brand decisions or commercial strategy are put into tools without thought.’’
With roughly a dozen tools in play, “our approach is to make approved tools useful enough that people don’t need to go around them, while keeping clear rules around what can and cannot be shared,’’ Jankie says. “AI tool sprawl is real because every function now has five new tools promising to save time.”
As a smaller company, the problem is not just cost, he points out, but “context fragmentation. If ideas, customer insights, creative drafts and operational decisions are spread across too many tools, you lose institutional memory.”
The company also takes a practical view to avoid opening the company up to risk. “We limit sensitive data exposure, prefer tools with clear business use and document repeatable workflows,” Jankie says.
It also separates experimentation from operational use.
“It’s fine to test a new AI tool, but it should earn its place before it becomes part of the workflow,’’ he says. “If it touches customer data, finance, legal, marketplace accounts or brand-critical work, the bar is higher.”
Understanding What AI Tools Can Do
Cutting down on the number of AI tools brands use requires knowing what AI can actually do compared to what you are hoping it can do, says Sawyer’s Loftin.
Calder recommends that companies refrain from viewing tool sprawl as a controlling challenge, but instead see it as a way to engage with the business regarding opportunity.
Leveraging trusted external partners can help shape a tool strategy and provide an extra source of support.
“People are just trying to do their jobs at the end of the day — be part of the solution,” says Calder.
The first rule is to manage use cases, not logos, says Natpat’s Jankie. “Companies can waste a lot of time debating which AI tool is ‘best,’ when the better question is: 'What job are we hiring this tool to do?'”
He recommends creating a short list of approved tools for common work, leaving room for controlled experimentation. Also, companies should define what data should never be pasted into external tools, and make outputs traceable.
“If AI helped create an ad, a support reply or a marketplace change, the final human owner should still be clear." says Jankie.
Further, he suggests reviewing the stack regularly. If two tools do the same thing, consolidate, Jankie stresses. “If a tool is exciting but not changing outcomes, remove it.”
The brands that win won’t be the ones with the most AI tools or the strictest bans, notes Bain’s Kömm. “They’ll be the ones with … a sanctioned platform people actually use, and the discipline to redesign work so AI compounds,’’ he says. “Sprawl is optional, scale is not.”
