Automate a process.
Stay in control of decisions.

An agent can find information, draft a response or suggest an action in your tools. We start with a specific process, compare the available options and define what should remain under human control.

Describe my process

We build custom AI agents for your business

What’s your business?Select

An agent to prepare your tender responses.

Your data

The tender documents

Your references, CVs and certificates

The agent checks the documents

Identifies requirements, finds the documents you have and asks their owners for anything missing.

In your tools

A submission ready for review

Shared files : Documents organised with source links
Task tracking : Missing documents assigned to owners
Email : Reminders tracked until documents arrive

A certificate has expired? The agent requests the latest version and updates the submission before your review.

Indicative estimates4 cases / month

less manual handling time: 25-38%
freed up per month: 8-12h

See the assumptions

Document retrieval, assembly and follow-ups. Excludes bid strategy and the decision to tender.

Assumed monthly volume: 4 cases / month
Manual time before / case: 480 min
Manual time with the agent / case: 300-360 min
Including human review: 60 min

Simulation after implementation, with accessible data and a stable process. Review and routine corrections included. Integration work and exceptional incidents excluded. Validate against your own data; results are not guaranteed.

Hours freed up = volume × time saved per case. Both figures express the same saving and must not be added together.

You stay in control.

Access, quality and costs are defined from the start. Controls are part of the product.

Limited access

Approved sources, role-based permissions and separate accounts. We define which data can be accessed, by whom and for what purpose.

Approved actions

Sending, editing or making a sensitive decision: the business owner approves the agreed actions. The agent can stop and hand over when information is missing.

Evaluated quality

Representative test cases, accessible sources, analysed errors and acceptance thresholds agreed together. A plausible answer is not proof of accuracy.

Visible costs and incidents

Cost per use, caps, alerts and logs suited to the data. Procedures explain how to pause, fix and resume the service.

A starting point
for every project.

From scoping to a live product, we join your project at its current stage. Each engagement has a scope, deliverables and a tailored proposal.

Already have scoping or a pilot? We build on what you have.

Find the right starting point

AI scoping

For a team deciding where to focus.

Decide whether an existing tool, automation or AI actually solves the problem.

Around 1 week · depending on scope and access

Interviews and a baseline for the process: volumes, time and errors to measure.
Comparison of options, data requirements, risks and estimated usage costs.
Scope, success criteria and a go / no-go decision in a reusable report.

Scope: One team, one business process and one decision-maker. The report remains useful if you choose another team for the next phase.
Prerequisites: Access to the people involved, a description of the tools and authorised data samples.
Out of scope: No development, company-wide audit or production deployment.

Scope my process

Measured AI pilot

For a team with a defined use case and criteria.

Test the value of a workflow before committing to production.

3–5 weeks · depending on scope and access

A testable workflow in a controlled environment, with a test group.
An evaluation dataset, observed errors and quality, cost per use and user feedback.
Demo, documentation and a recommendation to proceed to production or stop.

Scope: One workflow, one main data source and one standard integration, named in the proposal. Sensitive actions require human approval.
Prerequisites: Scoping or equivalent preparation, authorised representative data, test access and available users.
Out of scope: No production service with an SLA, complete migration or connections to an unlimited number of tools.

Prepare a pilot

Deployment and adoption

For a team with a validated use case.

Integrate the workflow into operations, with clear responsibilities and limits.

6–10 weeks · depending on scope and access

Agreed integrations, roles, permissions, error handling and logging.
Defined evaluations, spending caps, alerts and human approvals.
Documentation, recovery and shutdown procedures, training and handover to your teams.

Scope: Production deployment of the approved scope, acceptance testing and launch support. Integrations and acceptance criteria are listed in the proposal.
Prerequisites: A successful pilot or equivalent evidence, a business owner, production access and decisions on data handling.
Out of scope: Ongoing operations, on-call cover, guaranteed availability and unplanned features require separate contracts or work packages.

Prepare deployment

After launch

Operations defined at the right level.

Define the support you need

Optional support. Scope, hours, channels, incident tracking and response times are agreed in the contract. On-call cover and guaranteed availability are not included by default.

Separate costs. Models, cloud services and licences are listed separately. Incident handling is distinct from new features, which are quoted as separate work packages. Neither AI usage nor maintenance is unlimited.

Accounts in your name. Access, code and documentation can be transferred under the contract. We plan for data exports, vendor dependencies and handover or shutdown arrangements.

Our method

Each milestone,
a decision on what comes next.

Each phase ends with verifiable evidence and a shared decision. Deployment only begins once the necessary conditions have been approved.

Scope
Identify the starting point, options and prerequisites. Approve a scope or decide not to proceed.
Evaluate
Test the workflow with a user group, review its errors and costs, then decide whether to move to production.
Deploy and adopt
Approve access, acceptance testing, recovery procedures, training and support terms.

FAQ

How do you prepare a quote?

We start with the need, scope, access and integrations. The proposal separates the phases you need, development and model, cloud or licence costs. Existing scoping or a completed pilot can provide the starting point. For product development, the quote is tailored to pages, content and features, from a brochure website to a business app.

What are the indicative timelines?

Scoping: around 1 week; pilot: 3–5 weeks; deployment: 6–10 weeks. These are indicative timelines for the AI engagements described, once data, access and contacts are available. A website, app or platform follows a project-specific schedule defined in the proposal.

Do we have to start with scoping and then a pilot?

No. Existing scoping, a tested pilot or equivalent evidence can provide a starting point. Together, we review what is already established and what still needs validation. The scoping report can be used by your own team or another provider.

Are support and usage costs included?

Subscriptions, models and cloud costs are separate from development. An optional support agreement can define covered incidents, hours, response times and handover arrangements. Enhancements are quoted separately. On-call cover, guaranteed availability and unlimited usage are not included by default.

Which AI use cases are most useful in a business?

The strongest use cases fit into daily operations: internal agents, business copilots, retrieval-augmented search, document summaries, workflow automation, message classification and data extraction.

Do you build AI agents or RAG systems?

Yes. We design AI agents, internal assistants and RAG systems connected to your data and tools. Permissions, sources, human approvals, costs and safeguards are defined at the architecture stage.

How do you stop an AI project from remaining just a demo?

We start with everyday friction, validate a concrete use case, then integrate the system into existing workflows. Adoption depends on the interface, training, documentation, measurement and usage rules.

How do you handle security and data in an AI project?

We limit data collection, define access, separate environments, protect keys, document usage and favour architectures that keep sensitive data under control.

AI & Automation
Shall we talk?

Tell us about the process, the tools you use and the difficulty you face. We will use that context to define the next steps.

Describe my process

Automate a process.
Stay in control of decisions.

We build custom AI agents for your business