Logica · 2026.1Reasoning · Live

Decisions with substance

Logica is the reasoning layer of Agenivo. Instead of a black-box answer, your teams see every thinking step, every source and every confidence score – without making the answer slower.

Book a reasoning walkthrough
100 % Lineage
Audit pro Antwort
EU AI Act
Ready
Sample decision
0%sicher

Customer K. asks: "Can I still cancel my contract?"

  1. 01 Identify contract type
  2. 02 Check applicable law (BGB § 312g)
  3. 03 Compute cancellation window
  4. 04 Phrase the recommendation
Empfehlung

Yes. With an online sign-up on 9 May the 14-day cancellation window applies – you have until 23 May.

BGB § 312g · Contract #4218 · CRM snapshot 09:42
Why explainability

Three stakeholders who will not accept "because the model said so"

AI answers without reasoning fail in three places at once. Logica delivers the substance each stakeholder group asks for.

  • Perspektive 01

    Board

    Why should we trust the AI here?

    Because every answer has a sealed reasoning path. You can spot-check at any time – so can the auditor.

    Beweis

    Audit trail per answer · confidence score · data lineage

  • Perspektive 02

    Regulator

    Which data did you use? Which model? Why?

    Full lineage per decision. Rule set, model card and knowledge snapshot are documented – compliant with the EU AI Act.

    Beweis

    Lineage export · risk class · model versioning

  • Perspektive 03

    End customer

    On what basis is the AI telling me this?

    End customers see the source directly: contract clause, knowledge article or internal guideline – always with a reference.

    Beweis

    Source citation · trust indicator · escalation path

Hybrid engine

Three inputs. One reasoned decision.

Logica combines deterministic rules, generative models and your knowledge base – and only releases an answer when all three line up.

Reasoning Engine · LiveEU AI Act · Ready
  • RulesDeterministic

    Your business rules, compliance constraints and escalation thresholds – versioned like code.

  • ModelGenerative

    Frontier or open-weight LLM, picked by routing per task and cost.

  • KnowledgeRetrieval

    Your documents, contracts, FAQs – semantically indexed and tenant-isolated.

Logic Core
Reasoning · confidence · lineage
Output

Decision with reasoning

  • Answer + confidence score
  • Reasoning path (steps with source)
  • Escalation signal on low confidence

Answers below the configured confidence threshold are not released – they are proposed and escalated to a colleague.

Konfidenz-Median
0,94
Reasoning-Pfad p95
< 2,0 s
Audit-Coverage
100 %
Datenresidenz
EU-Central
Showcase

One decision, unfolded

Real reasoning path from an insurance pilot customer. From intake to released answer in 1.8 seconds – every step traceable.

LGC-2026-09-#4218
Insurance · claims intake
Konfidenz 94%

My car was damaged last night in the car park. Am I covered, and what do I need to do now?

Reasoning-Pfad · 6 Schritte
  1. 01 Classify damage type
    rules.claim-taxonomy v3.1
  2. 02 Retrieve policy
    crm.policy #874-201
  3. 03 Check coverage scope
    docs.policy-2024.pdf · §4 partial cover
  4. 04 Determine deadlines
    rules.deadlines · 7-day reporting
  5. 05 Phrase recommendation
    llm.gpt-4o · tone: empathic
  6. 06 Escalation check (rule out hail damage)
    rules.escalation-matrix
Freigegebene Empfehlung

Your partial-cover insurance generally covers this damage. Please photograph the damage and report it within 7 days via the customer portal, or reply to this message with the photos – we take it from there.

Policy 874-201 · §4 partial cover · 7-day reporting
Impact in numbers

What reasoning shifts in production

Median values from live Logica use cases of the past two quarters. Measured across all verticals — no cosmetic smoothing.

  • 01
    0 %

    Confidence median

    Across every answer released without escalation.

  • 02
    < 0 s

    Reasoning path

    p95 latency including rule check, retrieval and LLM reasoning.

  • 03
    0 %

    Audit coverage

    Every answer with a sealed reasoning path and source lineage.

  • 04
    0 %

    Blind escalations

    Fewer context-less hand-offs – because source & confidence travel with the answer.

Values from 12 Logica pilots in insurance, banking and healthcare. We share the methodology and raw dataset on request.

Use domains

Three industries where Logica is live today

Logica is valuable wherever explainability is not "nice to have" but a precondition – for regulatory, commercial or ethical reasons.

  • Insurance

    Claims triage with legal references

    Use Case · First capture of claims with automatic coverage check and deadline calculation.

    Wirkung · Claims handlers see case + recommendation + sources before they start work – not after 30 minutes of searching.

    Auto rate
    62%
    Escalation rate
    −54%
  • Banking

    Pre-check loan applications with reasons

    Use Case · Structured pre-check of loan requests against credit rules and regulatory constraints.

    Wirkung · Applicants get a reasoned indication in minutes. On rejection: concrete improvement guidance.

    Response time
    < 4 min
    Complaints
    −38%
  • Healthcare

    Patient triage in practice

    Use Case · First intake guided by clinical guidelines with traceable risk classification.

    Wirkung · Relieves doctors, catches emergencies, but every recommendation comes with reasoning and source.

    Documentation
    −71% time
    Triage accuracy
    96%
From use case to release

Four steps. Real reasoning. No theater.

Logica is not installed as a pilot and forgotten. We build the reasoning path together with your domain and compliance leads – and only go live once the confidence threshold demonstrably holds.

Time-to-reasoning
6 weeksGo-Live
  • 01
    Week 101/04

    Use-case audit

    We assess your decision use case, identify regulatory constraints, source systems and the confidence threshold you really need.

    Deliverables

    • Risk classification per EU AI Act
    • Source inventory (data, rules, knowledge)
    • Per-answer-type confidence baseline
  • 02
    Week 2–302/04

    Rules & knowledge

    Your business rules are captured in the Logica Studio as versioned blocks. Knowledge base is semantically indexed with strict tenant isolation.

    Deliverables

    • Rule set in TypeScript or visual editor
    • Vector index with source lineage
    • Escalation matrix with thresholds
  • 03
    Week 4–503/04

    Reasoning pilot

    Shadow mode on real traffic. Domain experts sample reasoning paths, compliance audits lineage – the model is tuned on your actual data.

    Deliverables

    • Shadow run on live data · 2 weeks
    • Reviewer console for domain & compliance
    • A/B comparison against status quo
  • 04
    Week 604/04

    Go-live with threshold

    Logica goes live – but only answers above your confidence threshold are released. Everything else is proposed and escalated. You stay in control.

    Deliverables

    • Staged rollout: 10 % → 50 % → 100 %
    • Live dashboard for confidence & escalations
    • Weekly lineage export for audit
FAQ

What reasoning means in practice

The most common questions about Logica in technical and legal context.

  • Regular AI returns an answer. Logica returns an answer plus reasoning path, source, confidence score and the condition under which it escalates. Without these four parts, Logica does not release an answer.

Next step

Watch a live decision with real reasoning

40-minute walkthrough: you bring a real question, we build the reasoning path together and you see each step with source and confidence. Afterwards you know whether Logica fits.