Agents connected to your data, not to a script
AI Integrations for Business
We integrate language models into your systems so the agent answers with the same information your team sees, rather than whatever it managed to memorise from a document.
Included
What this service covers
The difference between a chatbot and a useful agent is access. A chatbot repeats prepared answers; an agent checks your inventory, your calendar and your CRM before replying, and can act on them. That forces you to solve what almost nobody solves first: what the agent may touch, under whose permissions, and how it gets logged. That’s how it’s built here, and it’s the same thing already running in LabsChats and Propqix.
- An agent connected to your inventory, your calendar and your CRM
- WhatsApp, Instagram and Facebook handled from a single inbox
- Per-tool permissions: what it may read and what it may execute
- Handover to a person the moment the conversation calls for it
- A log of every action the agent takes, reviewable afterwards
- Your internal documents as a source of answers, without exposing them
What it’s for
What changes in your operation
- 01
It answers with your data, not with a guess
The agent checks the source before replying. If the price changed this morning, it answers with this morning's price.
- 02
Support does not stop at night
The questions that always repeat are resolved at any hour, and your team only gets what genuinely needs a person.
- 03
It can be audited
Every query and every action is recorded. When something goes wrong you can see exactly what happened and correct it.
- 04
It starts narrow
The first agent covers one concrete, measurable case. It grows once that case works, and not before.
The costliest mistake
Deploying an agent that answers nicely and knows nothing
An agent that never checks your stock or your calendar can only repeat what somebody typed into it. It answers fast, answers wrong, and the customer ends up asking for a human anyway. The difference between a chatbot and a useful agent is not the model — it is what it is connected to.
- What usually happens
- A document gets loaded into it and it is expected to guess the rest.
- What we do instead
- It checks the source before answering, and hands anything outside its remit to a person.
How the work goes
From the first call to production
The same process on every project, whatever the size.
- 01
What repeats
We read the real conversations of the last few weeks to learn what gets asked, how often, and what is answered today.
- 02
Connect the data
We define which systems the agent reaches, with what permissions, and what it must never touch.
- 03
Test on real cases
It is rehearsed against your own conversations, the awkward ones included, before it talks to a customer.
- 04
Launch and tune
It goes live on one channel first, we read the first week of conversations and correct with that in front of us.
When it applies
Situations this service solves
If one of these sounds familiar, this is probably the service you’re after.
You answer the same five questions all day
Price, stock, hours, location and payment methods resolve themselves, and stay on record in case the customer comes back.
You lose sales by replying late
Someone who writes at eleven at night gets an answer at eleven at night, and the appointment is booked before they look elsewhere.
You have years of documents nobody consults
Manuals, contracts and policies become searchable in plain language, for your team and without publishing anything outward.
Plans
What it costs
Prices in US dollars, with the scope written down. If your case fits none of them, that is what the last plan is for.
Basic agent
Answers what always gets asked, at any hour, on one channel.
$890one-time
+ $99 per month to run it
or $990 a year, two months off
Includes · 6
- One channel: WhatsApp, Instagram or the site
- Answers with your prices, hours and policies
- Script tuned against up to 40 real questions from your business
- Handover to a person when needed
- A panel to read and audit conversations
- One round of tuning in the first month
Connected agent
Checks your data before answering, across all three channels.
$3,200one-time
+ $220 per month to run it
or $2,200 a year, two months off
Includes · 6
- WhatsApp, Instagram and Facebook in one inbox
- Connected to stock, calendar or CRM
- Per-tool permissions: what it may read and what it may not
- Conversation tagging and follow-up
- Monthly tuning based on what was actually asked
- Monthly report of what it resolved and what it escalated
Agent with actions
It does more than answer: it books, quotes and records in your systems.
$8,500one-time
+ $560 per month to run it
or $5,600 a year, two months off
Includes · 6
- Executes actions: booking, quoting, creating orders
- Auditable log of every action taken
- Knowledge base built from your internal documents
- Escalation rules by type of case
- Test environment for changing the agent without risk
- Fortnightly conversation review for the first quarter
Custom
Several agents, several languages, or integration with your own product.
Based on scope
Includes · 4
- Several agents with different roles
- Support in several languages
- Model deployed on your own infrastructure
- Written availability agreement
The AI model usage is billed in your own OpenAI or Anthropic account, with no markup from us. The monthly fee covers running it, tuning and monitoring.
Seasonal pricing, subject to change. The price in your proposal is fixed from the moment you receive it. In USD, local taxes not included.
Also
Services that usually go together
Next step
Which of these do you need?
If you’re not sure, that’s fine too: the first call defines the scope before anyone talks about price.