Buying managed AI: questions to ask before you sign

How pricing works, and what it doesn't automatically include

Ask any provider for a complete written cost breakdown: setup, ongoing management, included software, usage limits and any changes at higher volume. Providers package these differently. Do not assume third-party fees are included or excluded merely because the service has a monthly price.

For Hampala, use the agreed scope and final price as the authoritative answer. For another provider, request the same clarity. If extra subscriptions or usage charges are proposed, establish who approves them and how they appear on the bill before connecting accounts. The buyer's goal is a predictable, explained cost, not a guessed industry standard.

It's also worth asking what happens if usage grows — if an AI agent handles ten times the volume of customer inquiries three months in, does the fee change, and does that trigger additional third-party usage costs that weren't part of the original estimate?

Check Hampala’s starting price and scope →

Why scope has to be agreed before work starts

The single most useful protection a small business has when buying a managed AI service is a written scope of work agreed before the provider starts, not after. Scope should spell out which workflows or tasks the provider is taking over, which tools they'll connect to, how changes or new requests are handled once the engagement is underway, and what's explicitly excluded.

Without that, it's easy for an engagement to drift in either direction: the business assumes the provider will handle something that was never discussed, or the provider treats a reasonable request as 'out of scope' and tries to charge extra for it. A defined scope, agreed before the first invoice, gives both sides something concrete to point back to when those disagreements come up.

This matters more with AI-specific services than with traditional outsourced work because AI workflows tend to touch sensitive systems — customer data, email accounts, payment tools — and because the technology itself is still new enough that provider and client often don't share the same assumptions about what 'automating this' actually involves. Getting the scope in writing forces both sides to work through those assumptions upfront rather than discovering the gap mid-engagement.

A practical checklist before signing

Before signing a contract with any managed AI provider, a small business should work through a short set of questions covering scope, cost, and data access. This isn't about distrust of any particular vendor — it's standard due diligence for any service where an outside party gets operational access to business systems and ongoing billing authority.

We've put together a supplier evaluation checklist covering the questions worth asking before engaging a provider like Hampala or any comparable managed AI service — available as a downloadable Word document or PDF for reference during vendor calls.

• Get the full list of third-party tools and subscription fees in writing, separate from the provider's base monthly rate
• Confirm what happens to pricing if usage volume increases substantially
• Require a written scope of work before any setup or integration work begins
• Clarify who owns the data, workflows, and configurations if the engagement ends
• Ask what access the provider has to business accounts (email, CRM, payment tools) and how that access is removed at termination
• Check how change requests outside the original scope are priced and approved

Further reference: NIST: AI Risk Management Framework

Ask for proof of the exception path

Request a demonstration with a missing field, a duplicate event and a request outside scope. Ask who receives the exception, how quickly it is reviewed under the agreement and what happens if that person is unavailable. A beautiful normal-path demo does not prove that a workflow will behave responsibly when it cannot complete a task.

Agree who can stop the system and how access is revoked if the engagement ends. Ask whether records and configurations can be exported in a usable format. Do not settle for a vague assurance that the provider “handles everything.” A dependable service should be able to explain both its responsibilities and its boundaries.

Use a small acceptance test before expanding

Choose one workflow and define representative examples, expected outcomes and prohibited actions. Record the baseline and include review effort in the evaluation. Approval should depend on the agreed evidence, not on how long setup took. If important data-handling questions remain unanswered, keep sensitive information out of the pilot.

The free Hampala supplier checklist covers scope, cost, access, retention questions, change control and exit arrangements. It is a practical discussion aid, not legal advice or a certification of any provider. Use it during the conversation, write down the answers and resolve gaps before granting operational access.