You probably don't need more AI. You need the right things to work better. The AI agents for business we build at Grippia each do one specific job and do it well: answering the routine questions, finding an answer in your documents, drafting the first version of a proposal, helping a technician in the middle of a repair. They work with your information, follow your rules and pass to a person whatever genuinely needs a person.
What an AI agent is, without the hype
An AI agent is software that understands what it's asked, looks things up in your company's information and does something with it: answers, searches, drafts, alerts. Unlike a form or a menu of buttons, nobody has to pick the right option. You write to it the way you'd write to a colleague.
What it isn't: a generic chatbot let loose on your website, answering anything. If AI doesn't improve the solution, we don't add it. And when an agent doesn't know something, it says so and hands it on.
Four agents we know how to build
The first is already a product of ours, Grippia Recepción. The other three are adapted to each company on the same engines.
Reception: a WhatsApp AI assistant for business that understands what people write
Let the routine questions answer themselves. It handles WhatsApp, knows your business rules and only passes on what needs your attention: bookings, appointments, incidents, the same question for the tenth time. It's the agent you notice on day one, because you see it on your phone the moment you message it.
It has its own product: Grippia Recepción, currently dressed for padel clubs. The same engine adapts to appointments, incidents or bookings in other businesses.
Company knowledge: ask your company what it already knows
The information lives in Google Drive, SharePoint, email, shared folders and the heads of two people. This agent answers from your own documents and tells you which document and which page the answer came from. Without that, nobody trusts it, and a tool nobody trusts gets abandoned fast. It fits professional services firms particularly well.
Proposals and tenders: from the brief to a first draft
Start from a draft, not a blank page. The agent breaks down the tender or RFP however long it is, extracts requirements, deadlines and scoring criteria, drafts the response section by section and flags what's missing. Your team stops reading hundreds of pages and starts deciding.
Field assistant: the experienced technician in everyone's pocket
For people working on the road, on site or up a ladder. They ask from their phone and get the answer from the manual, in their language, along with similar faults that have already been fixed. Useful when everything the company knows depends on the two people who've been there longest.
How we put an agent to work
We're based in Barcelona and we always start the same way. You tell us what's annoying you, not which AI you want. We look at people, tools, information and workflow. You get a proposal: what we'll build, when, and for how much. Fixed, in writing. We adapt the engine, ship the smallest useful version, and real people test it before it goes live.
We don't start from scratch: our conversation and knowledge engines have been running for years. That's why it's weeks, not months. More on how we work.
An agent works better on top of a layer
An agent that answers on WhatsApp needs to know which bookings exist, who's writing and what they've been invoiced. That's why the first step is often not the agent but systems integration: connecting the tools you already use so the agent reads live information, not a copy from last month. In a hotel or holiday rental, for example, the agent answers "what time is check-in?" from the actual booking. More in software and AI for hotels and holiday rentals.
A real case: Arketip's method
Alejandro is a coach, and his method used to live in scattered files.
"My method lived in an Excel and a stack of PowerPoints, and before every session I searched them by hand. Now I open the session and the script is already there, and new clients book their first conversation on the website, not over email."
It's what the knowledge agent does, at the scale of one person: what you already know, ready when you need it.
What an agent doesn't do
It doesn't decide where judgement matters. It doesn't answer what it doesn't know. It doesn't replace your team: it takes the repetitive work away so people have time to think, decide and talk to customers, which is everything software is bad at. Software searches, organises and suggests. People review, decide and act.
Where your agents live and what happens to your data
Usually we host and maintain the system for you, for a monthly fee and always on your own domain. Your documents and conversations are yours: if you ever leave, you take them with you. Each agent is a module on the same layer, so you can start with one and add the next once the first is working.