Questions Before You Book a Call
Grouped by what stage of the conversation it belongs to. If yours isn't here, ask it directly.
Getting Started
What is an AI agent, in plain terms?
Software that can read information, decide what to do with it inside rules you set, and then actually do it in one of your systems — not just describe what should happen. See what is an AI agent for the longer answer.
How is this different from a chatbot?
A chatbot answers questions from a knowledge base and stops there. An agent is connected to your actual systems and takes the next step — updates a record, sends a reply, files a document — inside rules you set.
Do we need to already have an AI strategy?
No. Most clients start from a specific, annoying, repetitive task rather than a strategy document. We'd rather scope one real task well than build a platform for a strategy that hasn't been tested yet.
What if we're not sure an agent is even the right fix?
Say that on the first call. Telling you an agent isn't the right fix — sometimes a simpler rule-based automation covers it — is a normal outcome of a discovery call, not a failure of one.
How Agents Work
What happens when the agent is not sure?
It escalates. Every agent we build has explicit boundaries for what it will act on alone and what it hands to a person — with the context already attached, not a blank ticket.
Can an agent make mistakes?
Yes, and pretending otherwise would be dishonest. The engineering work is in narrowing where mistakes can happen, catching the ones that do, and making sure a wrong action is easy to spot and reverse rather than silent.
Which AI model does the agent use?
Whichever fits the task and your existing stack — we're not tied to one provider. See AI agent engineering for how we choose.
Can one agent hand off to another?
Yes — that's a multi-agent system, where specialised agents pass work to each other rather than one agent trying to do everything. See multi-agent systems.
Integrations & Security
Can an agent connect to our existing tools?
Very likely, if it has an API or an interface an agent can act through safely. See AI integrations for the categories we connect to most often.
How much access does the agent get?
Only what the task needs, scoped through credentials you create and control — never blanket admin access as a default.
Where does our data go?
It's processed by the agent and the underlying model provider to do the job, and nowhere else by us. We'll name the specific systems and providers involved in your scope document before you commit to anything.
Can the agent be turned off instantly if something looks wrong?
Yes. Every agent we build includes a straightforward way to pause it, and access is granted through credentials you control, so you can also cut access directly at any time.
Pricing & Engagement
What does an agent cost?
It depends on how many systems it touches and how much judgement it needs to exercise, not a flat per-seat fee. Ranges and what drives them are on the pricing page.
How long does a build take?
A single, well-scoped agent is typically weeks, not months — the exact figure depends on how many systems it needs to reach and how much of the decision logic is already clear.
Do you offer a fixed price upfront?
Not before discovery — a firm number before that conversation is either padding heavily or planning to ask for more later. You get an indicative range immediately and a written quote after scoping.
Support & Maintenance
Do you support the agent after launch?
Yes. Systems and requirements change, and an agent that isn't maintained drifts out of date quietly. Ongoing support is part of the engagement, covered on the pricing page.
What happens if the system the agent connects to changes?
That's exactly what ongoing support covers — monitoring for breakage and adjusting the agent when a connected system changes shape, rather than leaving it to fail quietly.
Can we bring the agent in-house later?
That's a conversation we're glad to have — what a clean handover looks like depends on how your team is set up to maintain it, and we'll be direct about what that involves.
Still Something Unanswered?
Send it over before a build starts rather than after. A specific question gets a specific written answer.