AI
AI managed services
Ongoing operation of AI-assisted workflows, with people in the loop where the risk requires it.
What is delivered
Workflow design with defined human review points
Measurement of output quality, not just volume handled
A clear procedure for when the system should not decide alone
Why do in-house projects stall?
Because the prototype is the easy part. What fails afterwards is the operation: who corrects it when the system is wrong, who reassesses when the data shifts, who answers the customer who got the wrong reply, and who documents all of it for the regulator. None of those questions is settled by a demo.
It also fails for want of an evaluation set. A team building without having first set aside real, already-decided cases cannot say whether the system improved or merely fitted the examples it saw, and the difference between those shows up in production.
In a managed service those responsibilities are assigned before starting, and the evaluation set is built in the first phase and kept separate. That is the main difference and it is worth more than any accuracy gain.
Which processes are good candidates?
Those with high volume, stable rules and an outcome verifiable within hours. Routing messages by subject, extracting document fields, classifying requests by type, and drafting answers to frequent questions with human approval before they go out.
The criterion that matters most is the cost of an error. If a wrong classification means a message reaches the wrong team and someone forwards it in two minutes, the error is cheap and visible and the process is a good candidate. If it means a customer receives a wrong decision about an entitlement, it is not.
The second criterion is having historical examples. A process running for two years with a record of what was decided has evaluation material; a new process does not, and starting by automating something never done manually is solving two unknown problems at once.
How is human oversight designed?
With an explicit rule on which cases require approval before going out, written by case type rather than by confidence threshold. Thresholds are useful and are not a defensible criterion on their own: a system can be very confident and very wrong, and the difference is not observable from outside.
With enough time for review to be real. If the design implies each reviewer approving one proposal every ten seconds, the oversight is nominal. Sizing the review layer by how long an honest review takes is what separates compliance from theatre.
And with known control cases seeded into the queue, measuring whether review is actually happening. It is the only way to detect drift towards rubber-stamping, which always happens and never announces itself.
Which regulatory obligations apply?
Risk classification, first. Most mid-market cases fall under limited risk, where the central obligation is transparency: the user has to know they are interacting with an automated system, and artificially generated content has to be identifiable as such.
Some cases rise to high risk, particularly where the system participates in decisions about employment, access to essential services, credit or the assessment of people. There, obligations on risk management, data quality, technical documentation, event logging and effective human oversight are added.
Add the General Data Protection Regulation where processing involves personal data, with particular attention to automated decision-making with significant effects, which has its own regime and is frequently confused with automation in general.
How do you know it is working?
By total process time and reopening rate, not by percentage of cases automated. Automating eighty per cent and reopening half is worse than automating forty and closing nearly all, and the second scenario looks worse on any dashboard measuring only automation.
We also measure the time supervision consumes. A system automating half the volume and requiring a review layer the size of the previous team has released no capacity, it has only changed the kind of work, and that only shows if someone counts the hours on both sides.
And braking signals: complaints clustered on one case type, reopening rate rising with automation, and human reviewers who have stopped disagreeing. Any of these leads to suspending automation for that case type while we investigate, rather than adjusting a global threshold and hoping.
What does it cost and when is the return?
For two or three processes, setup typically takes between eight and sixteen weeks: four to six describing the process and building the evaluation set, four to six building and tuning, and the rest running assisted before supervision is reduced.
The return rarely comes from headcount reduction and almost always from cycle time and released capacity. A process that took two days and now takes two hours changes what the business can promise, and that is worth more than the direct cost difference in most cases.
The recurring cost people forget is maintenance. Data changes, processes change, and a system that is not reassessed degrades silently. Budgeting zero for maintenance is the usual way to have an excellent project in year one and a problem in year three.
What happens if you decide to bring this in-house?
We hand over the evaluation set, the system documentation with its known limitations, the change log with measured effect, and the annotation guidelines. It is the material that makes the system operable by another team, and it is yours by contract from the start.
What is harder to transfer is accumulated judgement on borderline cases, and for that reason handover includes a shadowing period where your team operates and we observe, rather than reading the documentation together. Reading is not the same as knowing how to operate.
Where ai managed services can be run from
Not every delivery model suits every service. The table shows only those that make sense for this work, with the data residency position of each.
| Model | Where | When it makes sense | Personal data |
|---|---|---|---|
| Onshore Portugal | Lisbon, Porto, Braga, Coimbra, Aveiro, Faro, Funchal and Ponta Delgada | When data cannot leave the EEA, or when the end customer is Portuguese | Stay inside the EEA. No transfer. |
| Nearshore in Portugal | Lisbon and Porto, for foreign buyers | When you need a multilingual European base without incorporating | Stay inside the EEA. No transfer. |
| Global network | Uzbekistan, the Philippines, Poland, the Dominican Republic, Mexico, Colombia, Turkiye and Africa | When you need continuous cover, specific languages or the lowest cost | Poland is inside the EEA. The others require standard contractual clauses. |
The data column describes the applicable framework and is not legal advice. The detail is in international data transfers.
Related services
- Training dataCollection, annotation and review of model training data, including European Portuguese and the African variants.
- Agents and automationAutomation of repetitive processes and assisted agents, with supervision and defined limits.
- Evaluation and safetyModel output evaluation, adversarial testing and safety review, by native speakers.
- Document intelligenceExtraction and classification of information from documents, with human verification where it matters.
Frequently asked questions
Do we need data scientists?
For most mid-market cases, no. You need defined processes and assigned accountability.
Where do we start?
With triage and classification, where an error is cheap and visible, and on a process with history that can serve as evaluation.
Should there always be human review?
On cases affecting customer rights, yes, and that rule is written by case type rather than by confidence threshold.
What is the right metric?
Total process time and reopening rate, not percentage automated.
Does the AI Regulation apply?
It does, with obligations proportionate to risk. Transparency is the most common in mid-market cases.
Who answers when the system is wrong?
We do, with responsibility assigned by contract before the first case passes through the system.
How long does setup take?
Eight to sixteen weeks for two or three processes, with half that time in description and evaluation.
Is there a maintenance cost?
There is, and budgeting it at zero is how you get an excellent project in year one and a problem in year three.
Let us look at the numbers for your case
Tell us which processes you want to outsource, in which languages and at what volume. We come back with a euro estimate and an operating design, with no commitment.
We reply within 6 hours on working days. If you would rather write: info@corpshore.solutions