AI automation for SMEs is not a fashion product. It is removing a step you already know how to do, that repeats, and that shows when it fails — and, when it helps, using AI to classify, draft or route inside that flow.
It is not “putting intelligence” on top of a mess. If the process is not written down, AI freezes the error with better prose. If nobody looks at the result, it freezes the error in silence.
This page defines what it is and what it is not. The decision guide — when yes and when not — is when to automate a process. The commission offer is AI automation, inside the same custom software service.
The short answer
A clear trigger. “When the customer accepts…” The data already lives somewhere. Not in a chat. Not in “I remember”.
One action. Creates the row. Pings whoever invoices. Moves the order. One.
A place you see the failure. The same day. If the failure shows up at month-end, make it visible first. Then remove the click.
AI only where the data already exists. Classify a document. Draft a report. Route an email to the right queue. Not “think the month”.
If A and B already do the work and only the trip is missing, integrate. If the gap is the work, build the flow. The test is the same: one step, one line, one owner. Guide: integrate or build.
What it is not (and when not to automate)
It is not replacing the ERP or the whole fiscal house in one commission. A bridge is not a system replacement.
It is not an ownerless pilot that never reaches production. If nobody looks at the queue on Monday, you do not have automation: you have an error that reproduces itself.
It is not judgement dressed as a cable. If half the cases are exceptions, you need a rule or a person — not a model. Judgement does not chain.
Do not automate when the process is undecided and you use “AI” to avoid arguing the rule. That does not save money. That hides. When the step changes every week, write the rule first. Then remove the click.
Do not automate because “we need AI”. If the click is cheaper than the bridge — fewer than about ten times a week on your own count — you do not need an industry benchmark. Count yours.
Not Mobility, Glowapp or Nummo. They are not Wallion products of this service.
How this shows up at an SME table (Barcelona / Spain)
You do not need a logo case study. The pattern is the table: a founder still signing quotes, someone in admin keeping the spreadsheet that “almost works”, and a partner (tax firm, ERP house, the a3-style stack next door) quoting the next module when Excel stops holding.
In Barcelona, the metro area and the rest of Spain, that table usually asks for the same thing in different words: “make it happen by itself”. Sometimes the pain is real — an accepted quote that never creates work, a closed timesheet that never pings invoicing, a record that lives in one system and not the other. Sometimes the pain is “digitise” without a single line. Wallion talks to that profile often: one concrete flow, not “the whole company”; fixed price; Diagnosis before build. No invented clients or volumes. The pattern is the table, not a case study.
If the trigger lives in María’s head and the “depending on the customer” rule changes every Tuesday, it is not an automation project. It is a process conversation. If the data already sits in the CRM and only the trip to the warehouse is missing, it is.
Failure modes (ones you have seen, or will)
AI on top of a mess.
You buy an “assistant” with no written trigger. Three weeks later it drafts well… and the same step still fails. Better prose is not a flow.
Eternal pilot.
There is an n8n, a Zapier, a script. Nobody owns it. Nobody looks at the queue. Failure shows up at month-end. Visibility first. Then the cable.
Exceptions as the real work.
You automate the happy path. Half the cases leave it. The team still does the important work by hand — and now also maintains the bridge. That is not savings. That is debt with better marketing.
Replacing the ERP “with AI”.
A brief that promises to replace the fiscal house or the whole ERP is not one-flow automation. It is another product. If you need that, it is another scope and another price — not this definition.
Cable with no “this is in / this is out”.
“Connect everything to everything” is not a scope. It is a wish. Without a trigger-and-action sentence, the quote is a hope with a figure on it.
Comparing a flow to a brand product.
Mobility, Glowapp or Nummo are not this service. Comparing the studio’s automation offer to those brands is mixing tables.
These failures separate a step that stops hurting from a pilot that accumulates.
What Wallion does (and when not to buy it)
Wallion builds AI automations and integrations inside the same custom software service, at a fixed price. It starts with a Diagnosis. One internal tool / one flow after Diagnosis sits in the published build band. The licence, if Wallion keeps what it built running, is monthly.
The model in one line: one trigger, one action, a place you see the failure; AI only where the data already exists; written scope, closed number, a date you can point at as done. It fits when you can write the step in one line and you want Diagnosis → fixed price → licence — not a counter “while we invent the AI”.
Do not buy automation with Wallion when the process is unwritten and you want “AI to discover it”; when half the cases are judgement and nobody will decide the rule; when the brief is replacing the whole ERP in one jump; or when you already run a living backlog of integrations and need hands — that is reinforcement / staffing: say it, do not hide it inside an “AI pilot”.
If on the call you say “this is judgement, not a cable”, say it in the first conversation. It does not hide in the FAQ. Studio geo: /en/software/barcelona.
Public prices (Option D — no invented figures)
Only the numbers that live on the site:
- Diagnosis — €1,900. One workflow audit, prototype, fixed build quote. If you proceed, the Diagnosis is credited against the build.
- Build — €6,900–14,900. One internal tool: one flow, after Diagnosis (e.g. the bridge plus the panel that runs it). Not the price of a large app, an MVP or staffing — those are a written fixed price, not that band.
- Licence — from €390/month. Hosting, security updates, regulatory and API changes, a ceiling of small-change hours, business-hours support — as published on the offer.
Apps, MVPs and staffing: same model of scope + closed number, different figure, in writing. This page does not invent tariffs. Full offer: AI automation.
Decision checklist (before you commission)
Tick each point. If any stays blank, do not sign “AI automation”.
- Can the step be written in one line — trigger + action — or not?
- Does the data already live in a system, or in a chat / someone’s head?
- Does failure show the same day, or at month-end?
- Who looks at the queue on Monday? Is there a named owner?
- Are half the cases exceptions (judgement), or is the happy path the real work?
- Does the step run more than about ten times a week on your count — or is the click still cheaper?
- Are we integrating (A and B already do the work) or building (the gap is the work)?
- Is the scope one flow, or “the whole house / the ERP”?
- Are we comparing the internal-tool band (€6,900–14,900) or another written fixed price (app/MVP/staffing)?
- Are we asking for a Diagnosis with a scope sentence, or an ownerless pilot?
Without those answers, “add AI” is an adjective, not a commission.
Questions buyers actually ask
Is this an “AI” project or custom software?
Custom software. AI is a piece of the flow when it helps — classify, draft, route — not the whole product. The hub remains /en/software. Module landing: /en/software/ai-automation.
When should we not automate?
When the process is undecided, or half the cases are exceptions. That is judgement, not a cable. Full guide: when to automate a process.
Do you use our current tools?
Yes, if the data already lives there. Integrating is usually the first move; building, when the gap is the work. See integrate or build.
Does the €6,900–14,900 band cover an app or an MVP?
No. That band is one internal tool / one flow after Diagnosis. Apps, MVPs and staffing: written fixed price, not that table.
Is the €1,900 Diagnosis lost if we don’t proceed?
If you proceed, it is credited against the build. If you don’t, you paid for a diagnosis with a prototype and a written fixed quote — not an empty deck.
What does the licence from €390/month include?
What the offer publishes: hosting, security, regulatory and API changes, a ceiling of small changes, business-hours support. Bigger changes: a fixed quote in days.
Do you work only in Barcelona?
The studio is based in Barcelona. We work with SMEs in the city, the metro area and the rest of Spain. Geo: /en/software/barcelona.
Do Wallion Mobility, Glowapp or Nummo belong here?
No. They are not Wallion products of this service. This page is about AI automation and integration in the studio’s custom software — not those brands.
What this page is not
Not the offer (/en/software/ai-automation; hub /en/software). It does not invent tariffs — Option D only. It does not treat Mobility, Glowapp or Nummo as Wallion products. It is not a ranking of automation tools. It does not replace the guide when to automate: that decides; this defines.
Thirty minutes
If the blocker is “can this be automated or is it judgement?”, say so in the first conversation. Book a 30-minute call.