Deciding to implement a chatbot or AI agent is the easy part. The hard part comes next: finding the company that builds, integrates and maintains it without becoming a permanent headache. The market is saturated with vendors that promise a lot and deliver little. This guide gives you concrete criteria to separate serious professionals from opportunists.
The 7 questions that separate professionals from opportunists
Before signing anything, ask these questions and demand concrete answers:
- «Can you show me an active customer in my sector with real data?» A serious vendor has specific success cases and permissions to show them (with customers who approved). If they only show generic demos, bad sign.
- «What happens when the chatbot doesn’t know how to answer?» There should be a clear protocol: routing to human, lead capture for follow-up, logging for future improvement. If the answer is «it always knows», they’re lying.
- «How does it integrate with my CRM/ERP/calendar?» Ask for concrete examples of integrations similar to yours. Integrations are where most projects die.
- «Who is responsible when something fails?» There should be a single point of contact, clear SLA, documented response times. If the answer is vague, in production you’ll have problems.
- «How are responses trained/updated?» Verify whether you can edit, or whether you depend on the vendor for every change (which scales badly in costs).
- «Does it comply with GDPR and do you have a DPA?» It’s legally mandatory. Any serious vendor delivers a Data Processing Agreement (DPA) without friction.
- «Can I see the metrics dashboard?» Ask for real screenshots: volume, resolution rate, conversions, satisfaction. Without metrics there’s no continuous improvement.
Table: vendor types and when to choose each
| Type | Best for | Risks | Typical cost |
|---|---|---|---|
| International SaaS platform (Intercom, Drift) | Generic B2B/B2C cases | Limited customisation, costs scale fast | €500-3,000/month |
| Local digital agency with chatbot product | Quick sectoral implementation | Highly variable quality, dependency | €2,000-15,000 setup + maintenance |
| AI-specialised consultancy | Complex cases with integrations | Higher cost, longer timelines | €10,000-50,000 initial |
| Individual freelancer | POCs and very simple cases | No guarantees, no structured support | €1,500-8,000 one-off |
| Internal team with low-code platform | Companies with technical profile | Learning curve, functional limitations | Variable |
Red flags to avoid
«24-hour implementation». Serious integration with CRM, calendar and validations requires weeks. «Express implementations» are basic chatbots without real integration.
«Works with any business without configuration». A chatbot that serves a clinic the same way as a restaurant understands neither. Sector and customer specialisation is where the value lies.
Dramatically low prices. If monthly cost is 5x lower than competition, something’s wrong: either it’s an inferior product, or hidden costs (additional messages, extra integrations, premium support).
No written contract or with vague terms. Demand a contract with SLA, responsibilities, exit conditions and data ownership. If they resist, keep looking.
«We guarantee X% sales increase». Nobody can guarantee business results without deeply knowing your operation. Vague increase promises are marketing, not real commitment.
Evaluating vendors? At StaffDigital we work as an AI agent specialised consultancy with documented implementations in clinics, ecommerce, restaurants, dealerships and real estate. Meet our team or request a session where we show you real cases applicable to yours.
How to structure the selection process
Step 1: Define the use case first. Before talking to vendors, be clear about what concrete problem you want to solve, what metrics you want to move, what budget you have.
Step 2: Request comparable proposals. Same use case to 3-4 vendors. Compare deliverables, timelines, prices, similar cases. The difference between offers is usually revealing.
Step 3: Request a paid but bounded pilot/POC. Before the full project, a 1-2 month pilot with a real case demonstrates technical and process capability. Worth more than any paper proposal.
Step 4: Talk to the vendor’s current customers. Ask for references and call. Ask what surprised them (positive and negative), what they would change, whether they would contract again.
Frequently asked questions
How much should a serious chatbot/AI agent cost?
For an SMB with typical cases: initial setup €4,000-15,000 (depends on integrations), monthly maintenance €300-1,000. If they offer much less, something’s missing. Much more without justification, also.
Is local or international vendor better?
For SMBs in the UK/EU, local vendor with sector and GDPR experience adds more value than generic multinational. For corporations with international presence, giants may make sense.
How long does a well-done implementation take?
Functional basic chatbot: 3-5 weeks. AI agent with integrations: 6-12 weeks. If they promise less than 3 weeks, check what they actually deliver.
Conclusion
Choosing the right vendor for your chatbot or AI agent determines project success more than the chosen technology itself. Take time to evaluate, ask for real demos, demand documentation, validate with existing customers. A bad decision costs thousands of euros and months of delay. Contact us if you want to see how we work.