The Operational Memory Layer · Swiss-Engineered
Your company's most valuable knowledge
is not in documents.
It is in how your people work every day. Alpix installs silently on enterprise PCs and captures the cross-software workflows your ERP, CRM, and SOPs never see — encrypted at source, stored on your infrastructure, inaccessible to anyone, including us. Built in Switzerland for Swiss SMEs.
The Invisible Layer
Your processes are documented.
Your real work is invisible.
Companies document processes in SOPs, ERP, and CRM systems. But real work happens between systems — in Excel, CAD tools, browser tabs, email threads, and informal sequences that are never recorded. That invisible layer is where the majority of operational knowledge lives.
Opaque to management. Impossible to standardise.
Unavailable for automation. Until now.
KPMG Process Excellence Survey 2023 — only 30–40% of enterprise work is captured in system event logs
Real work happens between systems.
SOPs, ERP, and CRM document what is supposed to happen. The actual work — in Excel, CAD, browser tabs, email threads, and informal sequences — happens between them. None of it is recorded.
Tacit know-how leaves with senior staff.
A senior engineer preparing a regulatory submission, a credit analyst closing a file, an auditor reviewing evidence — each applies years of judgment that is mentored, not documented. When they leave, it leaves with them.
You cannot standardise what you cannot see.
Independent studies estimate only 30–40% of enterprise work activity is captured in system event logs. The other 60–70% is opaque to management, impossible to standardise, and unavailable for automation.
Workflow Reconstruction
What ERP records.
What actually happened.
A finance controller preparing a quarterly budget update. 49 minutes of work. The ERP recorded one update and one email. Alpix reconstructed the entire decision sequence behind it — the cross-software workflow that produced the result.
What the ERP event log recorded
One budget update. One email sent. The other 47 minutes — invisible.
What Alpix reconstructed
The full cross-software sequence — segmented, labelled, and aggregated without individual reconstruction.
This is not monitoring.
It is your operational memory layer.
Alpix captures behavioural metadata, never content. No keylogging, no screen recording, no document reading. Findings surface at the role and cohort level — individual reconstruction is not architecturally possible.
Budget file opened in Excel
Fixed cost sheet reviewed — 3 min 47 sec
Q4 forecast tab opened
Q3 actuals file opened for comparison
Both files open simultaneously — cross-reference behavior detected
Specific cell range modified following comparison
Validation email drafted in Outlook
Email sent
The Engagement Model
Three tiers.
One compounding asset.
Capture first. Analyse later. Build AI on it last. Year one earns the trust; Tier 2 delivers the ROI; Tier 3 is the long-term thesis. The dataset compounds the whole way.
Capture the operational memory.
A lightweight Windows agent deploys across enterprise PCs. From day one, it captures real cross-software workflows and encrypts them on your infrastructure. No analysis yet — just accumulation.
- Lightweight on-premise agent
- Cross-software workflow capture
- AES-256 encryption at source
- Role and team scoping
- Local admin console
- revFADP & works council documentation
- Dataset growth telemetry
Convert the dataset into decisions.
Once 60–90 days of capture have accumulated, the analytics layer unlocks: bottlenecks, deviations, rework loops, time waste, and automation opportunities — quantified against your own baseline.
- Bottleneck detection
- Process deviation analysis
- Cycle time, rework, and time-waste mapping
- Cross-role workflow comparison
- Operational expertise mapping
- Automation-readiness scoring
- Longitudinal trend analysis
AI trained on how you actually work.
AI-assisted execution grounded in your client's own operational dataset — not generic LLM training data. Workflow-aware copilots that have observed real sequences. The data foundation is built today.
- AI-assisted workflow execution
- Trained on the client's own dataset
- Workflow-aware copilots, not generic LLMs
- On-premise inference — no data egress
- Grounded in real operational sequences
- Available to Tier 2 clients on roadmap
Land and expandTier 1 earns the trust. Tier 2 delivers measurable ROI once the dataset has depth. Tier 3 is the long-term thesis — AI grounded in operational reality instead of public text. We do not sell Tier 2 before the dataset is ready, and Tier 3 ships only on dataset foundations Tier 1 has been building.
The Engagement Arc
Engagements run 3–5 years by design.
By month 36 the dataset is operationally irreplaceable. Retention beyond that point is itself the impact signal.
Build the dataset
Capture begins. The encrypted operational asset accumulates on your infrastructure. No analysis yet — just the operational memory of the organisation.
Activate the intelligence
The dataset has depth. Tier 2 unlocks bottlenecks, deviations, and automation candidates — quantified against your own baseline.
AI on your operational reality
Tier 3 enables AI workflow execution grounded in how your people actually work. The data foundation is built today — and it is irreplaceable.
Privacy by Architecture
Privacy-preserving
by architecture.
Data sovereignty is the selling point.
And in Switzerland, it is also the law.
Data flow architecture
Endpoint
Behavioral signal captured
AES-256 at Source
Encrypted before first write
On-Premise Store
Your infrastructure
Analytics Engine
Deployed on-premise
This boundary exists because no code in our product is capable of crossing it.
100% On-Premise
The Alpix agent installs entirely within your network perimeter. No cloud dependency. No SaaS pipeline. Compatible with air-gapped environments. Your servers, your environment, full stop.
AES-256 at Source
Captured events are encrypted on the endpoint, before any local storage. Keys are generated during deployment and held exclusively by the client. Alpix does not generate, hold, or transmit keys.
Zero Data Egress
No workflow data ever leaves your infrastructure — not for processing, not for model training, not for support, not for anything. The product contains no outbound transmission capability.
Zero-Knowledge by Design
Alpix engineers cannot access client data. No support backdoor, no telemetry, no diagnostic channel. A design constraint, not a policy constraint — and the architecture is reviewable.
revFADP & GDPR Aligned
Built for the revised Swiss Federal Act on Data Protection from the ground up. Works council templates, employee notification materials, DPIA template, and signed DPA included in the deployment package.
Full documentation
Request the security architecture review
Full technical documentation provided to IT and security teams during scoping.
The Boundary
What Alpix is not built to do.
These are not policy choices. They are design choices. Findings surface at the role and cohort level via privacy-preserving aggregation. Per-employee reconstruction would require modifications the agent does not permit and the architecture does not support — a constitutional requirement of the Swiss social model, not just a feature.
- Track individual productivity metrics
- Score employees against KPIs
- Surface individual behavior to managers
- Generate per-person performance reports
The AI-Readiness Argument
AI models are trained on
what companies write.
Not on how they work.
Foundation models — GPT, Claude, Gemini — are trained on text, documents, tickets, and code. None of them have ever observed a real cross-software enterprise workflow. They cannot optimise what they have never seen.
Switzerland's 590,000 SMEs remain largely excluded from enterprise AI. The barrier is not access to models — it is the absence of structured, domain-specific operational data. Capture first; analyse later; let AI ride on top. That is the only sequence that produces AI grounded in operational reality instead of public text.
That dataset cannot be purchased, synthesised, or reconstructed retroactively. It can only be captured over time — and only starting today.
A company that begins capturing today and a company that begins in two years do not have the same dataset. They never will.
How Alpix Compares
Every adjacent tool stops where the real work begins.
| Capability | Process Mining Celonis, Signavio | RPA / Task Mining UiPath | Monitoring Tools ActivTrak, Teramind | Consulting Manual mapping | ALPIX This product |
|---|---|---|---|---|---|
| Captures cross-software workflows | |||||
| Works where ERP logs stop | |||||
| On-premise, zero data egress | |||||
| Dataset compounds over time | |||||
| Non-surveillance positioning | |||||
| Automated continuous capture | |||||
| European data sovereignty |
Celonis and SAP Signavio reconstruct from ERP event logs only — they cannot see human-computer work. UiPath Task Capture is cloud-based, which is a structural disqualifier for revFADP-regulated Swiss industries. ActivTrak and Teramind frame as surveillance and produce no process intelligence. MS Copilot and ChatGPT are cloud, single-app, and have no cross-software workflow visibility.
Who Alpix Serves
Built for Swiss SMEs
with 50–500 employees.
Across the sectors Switzerland leads.
The Defensibility Argument
Time is the moat.
And it only runs one direction.
Workflow data is temporal, company-specific, role-specific, and software-specific. It cannot be purchased. It cannot be synthesised. It cannot be reconstructed retroactively from any system log that exists today.
A Swiss SME that begins capturing today and a Swiss SME that begins in two years do not have the same dataset. They never will — and the gap compounds.
Dataset intelligence depth over time
Pattern significance, anomaly detection confidence, and seasonal intelligence accumulate non-linearly
The dataset you build in year one is worth ten times what it costs.
The dataset you build in year five is not for sale at any price.
Dataset value grows non-linearly
Workflow patterns become statistically significant after 60–90 days. Bottleneck and rework detection improves with each additional month. Seasonal patterns require 12+ months. 3–5 year datasets become irreplaceable.
You cannot capture yesterday's workflows
When a senior engineer or analyst leaves before their workflows are captured, that operational knowledge is permanently lost. Tier 1 protects institutional memory from day one — no analysis required.
The dataset is yours alone
No benchmark, no industry average, no synthetic corpus can substitute for your firm's specific operational patterns. The dataset is company-specific, role-specific, software-specific — and impossible to recreate.
The Engagement Arc
Alpix engagements run 3–5 years by design.
Not because we lock you in. Because the asset is just accumulation in the first six months, useful by month twelve, and operationally irreplaceable by month thirty-six. Retention beyond that point is the impact signal.
The longer you stay, the more your dataset is yours alone.
That is the product.
Your capture window
t+0
Every day delayed is operational knowledge that cannot be recovered.
Start the Engagement
Your most valuable knowledge is in how your people work every day.
Alpix is the operational memory layer that Swiss companies — and the AI systems they will rely on — depend on.
The capture window opens the day you start.