Method
Two routes to bringing AI into your company, depending on the problem
At Xarxalia, a custom AI project moves through phases, with a pilot before production, and its duration is set in Discovery. MIA takes a shorter route: its digital employees are already built, and the target is to have them live within five working days of getting access to your systems and the information we need about your business.
Custom AI projects
Six phases, each with a clear outcome
Nothing goes into production until it has worked in a pilot on a real case. How long each project takes depends on its scope, and we agree it in Discovery.
01
Discovery
What we do
We map the process with the people who run it every day: what comes in, who decides, which systems are involved and where time is lost. We agree what the AI has to solve and how that will be measured.
Outcome
Agreement on the problem, the systems involved, the success criteria and the timeline for the following phases.
02
Architecture
What we do
We start from what the HUB already covers and decide what needs building, which systems the solution connects to and what each agent can do without asking.
Outcome
The solution design: components, integrations, data flow, roles, permissions and autonomy levels.
03
Pilot
What we do
We put the solution to work on a real, well-defined case, at the lowest autonomy level and under human oversight. We measure it against the criteria agreed in Discovery.
Outcome
A joint assessment of what worked and what didn't, and the decision to continue, adjust or stop.
04
Integration
What we do
We connect the solution to the organisation's systems, such as the ERP, CRM, Microsoft 365 or databases, via API, MCP or A2A. Users can sign in with corporate SSO, with permissions set for each role.
Outcome
Integrations built and tested against the real systems.
05
Production
What we do
The solution goes live for all intended users, with human oversight, every action logged and close monitoring at launch. Autonomy goes up when the organisation decides.
Outcome
The solution in production and the organisation's team ready to use it.
06
Evolution
What we do
We review the results with data, adjust the solution and extend it to other processes. If something is needed that the HUB doesn't do yet, we build it.
Outcome
Data-based tracking and a plan for the next improvements.
MIA · Roll-out in five working days
Five days to put MIA to work
Day 1 · Setup and channels
Day 2 · Knowledge
Day 3 · Training
Day 4 · Testing
Day 5 · Live
Day 1
01
Setup and channels
We set up the company, the users and the permissions, and connect the channels your customers come in through.
On the table
Day 2
02
Knowledge
MIA learns all it needs to know about your business. We reuse the website and brand work you've already done.
On the table
Day 3
03
Training
We define the personality, the tone, the rules and what happens when something goes off script.
On the table
Day 4
04
Testing
Real scenarios before it talks to any customer, including the ones that go wrong. If it's going to fail, let it fail with us.
On the table
Day 5
05
Live
It goes live at the lowest autonomy level, closely watched. We only raise it if you ask.
On the table
MIA · What comes next
The first three months, month by month
First month
Tuning
We correct the tone, add real cases to MIA's knowledge as they come up and adjust the escalations. First report with your own data.
Second month
Expanding
A second digital employee or a new channel is added. We start reading the metrics as a trend, not as one-off snapshots.
Third month
Scaling
Autonomy levels go up where trust has been earned. We review the return in figures and define the next processes.
MIA · The monthly report
Every month we tell you what happened
A document you can show to anyone, with the figures checked and what we recommend for the month ahead. If there's nothing to show one month, we tell you that too.
Conversations handled, and on which channel
Average response time
Opportunities spotted and recovered
Quotes followed up and closed
Escalations to a person, and why
What we recommend for next month
Support and incidents
When something breaks, we already know
Incidents are graded, have maximum response and resolution times, and serious ones end with a written account of what happened.
Portal, email, WhatsApp and phone
Incidents graded across four levels
Maximum response time per level
Written account of serious incidents
Quarterly service review
What problem should artificial intelligence solve in your company?
Before we talk about technology, we want to understand the problem. Tell us which process you want to transform, which systems are involved and what result you need.
Frequently asked questions
What exactly happens on day one?
On day one MIA isn't speaking to any customer yet. We set up the company, the users and the permissions, and connect the channels your customers come in through: WhatsApp, email, web and calendar. Your business knowledge goes in on day 2, before MIA answers anyone.
Do the five days start when I sign or when I give you access?
MIA's five days start once we have access and your business information, not when you sign. Xarxalia's target is five working days: setup and channels, knowledge, training, testing and go-live. On day 5 MIA goes live at the lowest autonomy level, closely watched.
What do you need from me to get this started?
Xarxalia needs access to your customer channels and material about your business. On day 2, MIA learns from your website and catalogues, prices and terms, FAQs and internal protocols. Day 3 covers tone of voice, business rules, escalation rules and approved replies. You decide when the autonomy level goes up.
If something breaks once it's live, who do I tell and when do I get an answer?
You report it through the portal, by email, on WhatsApp or by phone, and Xarxalia grades the incident at one of four levels. Each level has a maximum response time. If it's serious, you get a written account of what happened. On top of that, the service is reviewed every quarter.
How will I know whether MIA is actually working for me?
If MIA is helping you serve more customers, better, and win more business, it shows in Xarxalia's monthly report: conversations handled by channel, average response time, opportunities spotted and recovered, quotes followed up and closed, and escalations to a person with the reason. In month three we review the return in figures. If a month has nothing to show, the report says so.
What is decided in the Discovery phase of a custom AI project?
In Discovery, the first of the six phases of a custom AI project, Xarxalia and the organisation decide what the AI has to solve, how that will be measured and how long the following phases will take. All of that is agreed before any building starts. The length depends on how many processes the solution covers, how many systems it has to integrate with and how much needs building beyond what the HUB, Xarxalia's own platform, already does. Before Discovery we don't yet know the systems or the volume of work, so any timescale would be guesswork. The five-working-day target is for MIA only.
What happens if the pilot doesn't work?
If the pilot of a custom AI project doesn't meet the success criteria agreed in Discovery, Xarxalia sets out what went wrong and recommends adjusting the design, changing approach or stopping the project. The pilot is there to test the solution on a real, well-defined case before investing in integration and production. With those results in front of it, the organisation decides whether to go into production. One commitment Xarxalia publishes on its Technology and trust page is never to say a feature is ready when it isn't.