For a large central bank, choosing SupTech is a major programme with a team to run it. For most of the world's financial authorities it is harder: smaller budgets, slower procurement, few specialist staff, and a choice that will shape supervision for a decade.
Why the choice matters more for smaller authorities
CGAP's 2026 research found AI-powered SupTech in use at 85% of advanced-economy authorities but 53% in emerging markets, with limited IT capacity and analytics skills the most cited barriers. A Pacific regulators' report noted that smaller jurisdictions may not reach the critical mass to attract vendors at all, and that most still receive returns as spreadsheets by email. When resources are this tight, a platform that needs a large in-house team to configure and run it is the wrong platform, however good its feature list.
The eight criteria that matter
1. The data model. Does the platform store data as structured facts, with XBRL, Excel and CSV as inputs, or is it built around one document format? The first can grow into granular collection; the second needs replacing.
2. Rule ownership. Can your own team change a validation rule, versioned and auditable, without waiting for a vendor release?
3. Change speed. How long did the vendor's last client take to put a framework change live? Ask for a named example measured in weeks, not a roadmap.
4. AI with evidence. Where does AI sit: summarising documents beside the work, or reading returns inside it? Can every AI-assisted finding be traced to its data and rule?
5. Hosting and lock-in. Can the platform run in your environment or region, and what do you take with you if you leave: data, rules and history?
6. Delivery and skills. Who configures, extends and runs it? A vendor that embeds engineers with your supervisors closes the skills gap instead of widening it.
7. Incremental replacement. Can it replace the riskiest legacy piece first, rather than demanding a big-bang migration?
8. Total cost over five years. Licence, configuration, change requests and exit, not just year one.
Procurement that works at small scale
Long tenders favour long programmes. A fixed first phase with a single workflow live in weeks lets an authority prove value before committing further, and gives procurement a concrete result to evaluate. Shared or regional platforms are another route some smaller regulators are exploring; the same eight criteria apply.
- 1Is the core a structured data model, with XBRL, Excel and CSV as inputs?
- 2Can our team change validation rules without a vendor release?
- 3What is the vendor's fastest recent framework change, with a named client?
- 4Does AI read returns inside the workflow, with every finding traceable?
- 5Can it run in our environment or region?
- 6What do we keep if we leave?
- 7Who configures and extends it, and will they work inside our team?
- 8Can it replace one legacy piece at a time?
- 9What is the five-year cost, including change and exit?
- 10Which workflow will be live first, and when?
To see where your authority stands today, take the three-minute readiness check, or readwhat SupTech is and how XBRL works.