Your approved AI platforms sit idle while staff cobble together workarounds with personal ChatGPT accounts and scraped data feeds. These shadow systems now drive more daily decisions than anything the transformation office has rolled out.
Procurement still buys seats based on the official list. Meanwhile the real usage shows up in expense claims for API keys and forgotten subscriptions that never hit the central ledger.
Risk teams write policies that assume central control. The policies get ignored because they add friction without solving the actual workflow blockages people face.
When something goes wrong the organisation has no audit trail. A bad output from an unofficial model triggers rework or customer complaints, yet no one can trace who authorised the tool or the data it used.
Finance sees flat or rising tool costs but cannot link them to outcomes. The productivity claims in the annual report rest on activity that the official stack never captured.
Fixing this requires killing the gatekeeping model. Grant teams direct access to vetted models with usage logging, then shut down the shadow routes by making the official path faster than the workaround.
Anything less just pushes the leakage deeper underground until the next budget review discovers another surprise line item.