AI Subscription Sprawl: Why Teams Pay for Five Tools and Use None

AI subscription sprawl arises when teams, afraid of missing out, book their own AI tools without anyone overseeing cost and usage. The remedy is an inventory of all subscriptions, a decision per tool by usage and business result, and a shared selection with clear ownership. This cuts cost and creates the basis for turning trial licences into a tool with measurable benefit.

A pattern has appeared in many companies since the launch of ChatGPT. Sales tests a writing tool, marketing two image generators, development a coding assistant, and the managing director a chatbot on a monthly plan. Billing runs through company cards, private accounts or expense claims, and no department knows what another uses. This article describes why this happens, how you can record cost and usage within two weeks and how the sprawl becomes a steered selection of tools. It provides an overview and does not constitute legal advice.

Why does AI subscription sprawl arise?

AI subscription sprawl arises because tools can be booked in minutes while the hurdle for a company decision sits higher. A subscription for the common offerings costs around 20 US dollars a month, which is below the thresholds at which procurement gets involved. Whoever expects an advantage books it. Added to this is the pressure that executives and colleagues create when they hear of successes with AI, and the worry of falling behind competitors. How this pressure works is described in the article on AI activism.

The result is understandable and still expensive. Several teams buy tools with overlapping functions. Licences keep running although nobody uses them, because the trial month turned into an annual plan. Confidential documents end up in private accounts whose settings nobody knows. And nobody can say what benefit the company draws from the sum of the spending. Responsibility lies with no one, because every purchase was small on its own.

What do the common AI subscriptions cost?

The common AI subscriptions for individual users cost around 20 US dollars a month, and offerings for teams and companies cost around 25 to 30 US dollars per user and month. The following prices come from the providers and apply at the time of this article.

The sums grow quickly. With 40 employees holding three individual subscriptions of 20 US dollars each, the cost is 2,400 US dollars a month and 28,800 US dollars a year, without the company having settled contracts, data processing or support. A team offering with central administration costs only slightly more per user than an individual subscription and supplies control, invoicing to one point and settings for how company data is handled.

How do you record cost and usage? An inventory in two weeks

You record cost and usage in an inventory that is complete within two weeks and draws on four sources. Finance supplies all statements from company cards and expense claims that relate to AI tools. IT supplies logins through the company account. The teams report in a short survey which tools they use and for what. In addition you ask the providers for the number of active users in the last thirty days, as far as the admin console provides it.

The result is a table with one row per tool: provider, number of licences, monthly cost, active users, owner and use case. The survey needs a clear assurance from the executive team that it carries no sanction. Whoever expects punishment does not report, and the inventory stays patchy.

A worked example shows the effect. A company pays for 25 licences of a writing tool at 20 US dollars a month, which is 6,000 US dollars a year. Six people used it in the last thirty days, so utilisation is 24 per cent and idle spend 4,560 US dollars. That is money without return, and the figure repeats with every tool on the list.

Keep, merge or end? The decision per tool

The decision per tool follows four criteria: utilisation, benefit for a business result, overlap with other tools and handling of company data. They yield three routes.

The thresholds in the table are rules of thumb from practice and can be adjusted. What matters is that they are fixed before the review and apply equally to all tools. The decision then becomes factual, and no team has to defend itself against an individual decision by the executive team.

How do you introduce a shared selection?

You introduce a shared selection in four steps that are complete within a quarter. In the first step the executive team names a responsible person who assesses tools, concludes contracts and answers the teams' questions. In the second step the business units define the tasks for which they want to use AI and measure the baseline, for example the time for a report draft. In the third step the responsible person selects, with IT, one or two standard tools per field of work and books them centrally. In the fourth step the person trains the users on real tasks and checks after ninety days whether the metric has moved.

Room for experiments belongs to this. Teams may test new tools within a defined frame, for instance with a time-limited trial account and without confidential data. Whoever applies after a successful trial receives a vetted tool. This keeps curiosity alive while the organisation keeps its overview. How employees can work with Microsoft products is described in the article on Microsoft 365 Copilot.

When does an enterprise offering beat individual subscriptions?

An enterprise offering pays off as soon as more than about ten people use the same tool regularly or confidential data is processed. The reasons lie in central administration, billing to one point and the settings for handling data. The price difference per user is small. The decisive question is who is responsible for operations, support and training. This role must be named before the purchase, otherwise the same sprawl arises at a higher level.

There is no upper limit for the number of tools. As a rule of thumb, one standard tool for text work and another for programming are enough, plus specialist tools where a business unit proves a use case with a metric. Which questions executive teams should ask at the start of the year is described in the article on AI priorities for 2024.

Tool order with Niehues Consulting

Niehues Consulting is a consultancy for AI management and organisation design from a delivery perspective, based in Düsseldorf. Founder Marin Niehues brings AI into operations, in projects ranging from VC-funded scale-ups to DAX 40 corporations and in more than 20 countries on six continents. In the service area of AI management, the work runs from business analysis and requirements through the business case to the selection and sequencing of applications.

The approach follows the Niehues Delivery Model with its three phases. For ordering tools, three of its checking questions count: Which business result should change measurably? Who is responsible for operations from day one? Are the users enabled and the workflows adapted? The entry point is a free initial consultation of 30 minutes, which you can arrange via the contact page.