Indicative CHF pricing. Final scope, third-party costs and tax treatment are confirmed in the proposal. Care includes an agreed support allowance, not unlimited support.
WHAT SHAPES THE PROPOSAL
The range follows the work.
The work
Number of workflows, exceptions, approval decisions and required outputs.
The connections
Available interfaces, permissions, source quality and integration constraints.
The operating context
Security review, acceptance evidence, handover and support requirements.
NO HIDDEN ASSUMPTIONS
Confirm the complete cost.
Third-party subscriptions, model usage and platform fees are confirmed separately.
Tax treatment and payment terms are confirmed in the written proposal.
Data remediation, unsupported interfaces or wider organisational rollout may need separate scope.
Care includes a defined allowance, not unlimited support or an implied 24/7 service.
COMMON QUESTIONS
Price is only part of the decision.
Is AIFAB only for smaller companies?
No. We work with SMEs and large companies. The starting point is a defined workflow, an accountable owner and a measurable outcome. In larger organisations we also scope the relevant governance, access, procurement and integration requirements. The published offers describe bounded engagements, not an unlimited company-wide rollout.
What does an AI workflow cost?
AIFAB Workflow Launch is indicatively CHF 9,500–18,500 for one workflow, one team and up to approximately three systems; typically CHF 14,900. Team Launch is CHF 5,500–9,500. A complex workflow may start with a Workflow Blueprint at CHF 2,900. The proposal confirms scope, exclusions, third-party costs and tax treatment before work begins.
How long does implementation take?
Workflow Launch typically takes 10–15 business days after scope, access and sample data are ready. Team Launch takes 5–7 business days; an Operations Pack takes 4–6 weeks after its required Blueprint. Integration access and customer review can affect elapsed time.
Do we need to replace our current systems?
Usually not. We start with your existing inbox, CRM, spreadsheets and approved knowledge. We confirm API access, permissions and data quality during scoping. If a system cannot support a safe integration, we explain that before committing.
Who owns the implementation?
Production runs in customer-owned accounts, with customer-controlled credentials. Handover includes the agreed implementation, documentation and operating responsibilities. Third-party platforms remain subject to their own licences.
Where does the data go?
That depends on the agreed systems and model provider. We document the data path, approved sources, permissions and retention requirements during scoping. Public demonstrations on this site run locally with synthetic data and send no customer information to a model.
Can we stop it instantly?
Yes. Every implementation ships with the Red Button: a one-click pause held by a person you designate, typically the COO or the process owner. New automated actions stop, queued items keep their state and owner, and new cases follow the documented manual path until an authorised person restores operation. A pause cannot recall an email that has already been sent; the exact stop boundary is agreed per system during scoping.
Can the models run in Switzerland only?
Swiss hosting can be assessed through a Workflow Blueprint. The reference option is not yet a validated AIFAB delivery path. Hosting, storage, logging, support access, model-provider egress and equivalent approval controls must be verified before any residency commitment.
What happens when the AI is uncertain?
The workflow can pause, request missing information or route the item to a named person. Validation and business rules run around model steps. We agree uncertainty and failure cases before launch, rather than assuming every model output is usable.
Can messages be sent automatically?
Technically yes, when the connected system permits it. Initial implementations normally require human approval before external sending. Any exception needs an explicitly agreed authority boundary, suitable constraints and acceptance tests.
Can AIFAB work with Microsoft 365?
Microsoft 365 can be part of a scoped workflow, subject to your tenant policies, licences and administrator-approved access. We establish the minimum permissions needed before connecting Outlook, SharePoint or other services.
Can AIFAB work with Google Workspace?
Google Workspace can be part of a scoped implementation, subject to administrator approval, available APIs and your organisation’s data policies. Gmail, Drive and Sheets access is limited to the agreed workflow.
Can you connect Salesforce, HubSpot or custom systems?
We assess the available APIs, account permissions, rate limits and data model first. A named platform is not a guarantee that every feature or subscription tier can be integrated. Custom systems need a supported interface or a separately agreed integration.
Can you assess Bexio, Abacus or Odoo workflows?
Yes, as a scoped assessment. We first confirm the product edition, available interfaces, administrator-approved access, data model and intended write actions.
When is a Workflow Blueprint required?
A Blueprint is intended for complex or multi-system work and is required before an Operations Pack. It resolves the system, data, risk, authority and value questions needed for a fixed-price implementation decision. Its fee is credited in full when the implementation is signed within 60 days.
What happens after launch?
We train the users, document ownership and review the agreed outcome. AIFAB Care can provide monitoring, minor adjustments and a defined optimisation allowance. Support scope and response arrangements are agreed; support is not unlimited.
Which processes should not be automated?
Processes with unclear ownership, unreliable inputs or no repeatable decision path need attention first. We do not automate consequential legal, financial, employment or safety decisions by default. Sensitive or regulated use cases require separate expert review.
How is ROI measured?
Start with a baseline: volume, handling time, rework and the cost of operating the workflow. After launch, compare the same measures. Time released is capacity, not automatically cash saved; adoption, review effort and ongoing costs belong in the assessment.
Is an AI agent always necessary?
No. Deterministic validation, calculations and routing should stay deterministic. AI is useful where interpretation, classification, extraction or drafting adds value. We add autonomy only when it is needed and its authority can be controlled.
THE NEXT USEFUL STEP
A useful first conversation needs one workflow.
Bring its owner, the systems involved and an approximate volume. A redacted example is enough to start assessing fit.