Mortgage Pathways

Zeus AI Proposal

Prepared for Mark Finch and Stephen Mills, Mortgage Pathways · Zeus Client ID MP-2026-0324 · Date: 2026-03-24 · Rev B · Subject to copyright

A focused AI automation build to remove manual re-keying from your mortgage workflow by turning 3CX call transcripts into structured, system-ready updates.

1.2 Mortgage Pathways

Business Context

  • Current stack: OnceHub -> Microsoft 365 diary -> 3CX -> Toolbox / Mortgage Brain -> eKeeper -> lender portals.
  • Lead volumes are high: 225 booked, 203 attended, and around 400 monthly appointments including adviser-sourced flow.
  • Current completed volume is around 150 mortgages per month.
  • Main friction is repeated manual re-keying and admin overhead across systems.
  • Advisers feel overworked due to duplicated process steps per case.

Core Objective

  • Automate transcript-to-system updates from 3CX into Toolbox and eKeeper.
  • Generate lender-portal-ready copy blocks to reduce typing and speed completion.
  • Keep compliance and quality controls with validation before write-back.
  • Use Zapier only where simple; implement custom logic where process complexity requires maintainable code.
  • Finnova API + connected tools are part of the integration pathway.

Mark Finch

LinkedIn profile photo for Mark Finch

Operations and process lead

Leads operational process design and rollout priorities.

Stephen Mills

LinkedIn profile photo for Stephen Mills

Director and executive sponsor

Owns governance, delivery sign-off, and commercial decisions.

1.3 Zeus AI - Block Preview

Zeus AI Ltd

Zeus AI Ltd is a London-based technology company that specialises in setting up AI agents that nurture contacts and book calls for your team.

Registered address

71-75 Shelton Street, Covent Garden, London, WC2H 9JQ

VAT number

472247292

Company number

14908132

Senior management team

Portrait of Sam Oliver

Sam Oliver

CEO

  • Founded Lead.Pro and scaled to 37% of UK estate agents before exiting to Tosca Fund.
  • Network in real estate and financial services.
Portrait of Matea Radovan

Matea Radovan

COO

  • Scaled Lead.Pro revenue 9x to GBP 127,000 MRR.
  • Experience implementing new technology in large enterprises.
Portrait of Mladen Pecanac

Mladen Pecanac

CTO

  • 20+ years of experience leading engineering across startups and global enterprises.
  • Network of high-performing AI developers in Serbia.

2.1 How it works

This implementation focuses on one clear workflow: 3CX transcript intake -> structured extraction -> mapped write-back into Toolbox and eKeeper.

The reason this is delivered as custom automation (not only Zapier) is that your process requires field-level validation, cross-system reconciliation, and reliable retry/error controls.

Outcome target: remove duplicate typing, improve record consistency, and free adviser time for client conversations.

Core implementation sequence

1) Define extraction schema

Identify the exact fact-find and case fields to be extracted from transcript data.

Includes:

  • Client details and affordability-related fields
  • Mortgage preferences and rationale fields
  • Compliance-sensitive items requiring confidence checks
  • Lender copy-ready summary formatting outputs

You approve the schema before production mapping begins.

2) Build integration and validation layer

Connect ingestion, parsing, and mapped write-back logic with QA safeguards.

This includes API-first paths plus fallback handling where providers limit direct endpoints.

3) Deploy with operational controls

Run UAT, release to live, monitor errors, and hand over a practical operating model.

Optional phase two can add WhatsApp nurture once the core admin-efficiency workflow is stable.

2.2 Results

Transcript Automation Projected Results

Performance depends on transcript quality, mapping rules, and operational adoption by adviser teams.

This projection model shows:

  • the monthly appointment volume entering the process
  • the number of cases expected to reach structured call output
  • the estimated completed mortgages at current and improved conversion rates

For this proposal, we are using a value of per completed mortgage. This value is editable on each proposal.

Outcome summary

    ROI

    2.3 Timeline

    Timeline

    Total plan length: 30 working days - Day 1 of 30 - Current stage: Brief clarification and KPI definition

    Week 1 - Clarify brief, deliverables, and KPIs
    Today
    5d
    Week 2 - Validate 3CX transcript access (API + fallback route)
    5d
    Week 3 - Build extraction schema and parser environment
    5d
    Week 4 - Map and sync with Toolbox + eKeeper
    5d
    Week 5 - QA/UAT and lender copy-output hardening
    5d
    Week 6 - Go-live, hypercare, and handover
    5d

    Six-week delivery model with one-off build fee and optional post-launch maintenance.

    3.1 Quote

    Recommended

    Mortgage Pathways transcript automation build

    GBP 7,200 discounted one-off build fee (ex VAT)

    • Standard development rate: GBP 640/day (GBP 3,200/week / GBP 12,800 per 4-week month), ex VAT
    • Discounted implementation rate: GBP 1,200/week (GBP 4,800 per 4-week month), ex VAT
    • Discount applies where agreement is signed before financial year-end and due to potential alignment as a Pivotal-owned brand

    Maintenance and running cost (MRR)

    GBP 1,760/month standard fee (ex VAT)

    • Discounted MRR: GBP 860/month (ex VAT)
    • Base maintenance + performance tuning support
    • Variable usage and provider costs charged separately

    Commercial terms

    Payment split 50% kickoff / 50% go-live

    • Scope includes transcript agent + Toolbox/eKeeper sync + lender-ready copy output
    • Optional next phase: WhatsApp nurture journeys for developer/estate lead cohorts
    • All prices shown are ex VAT

    3.2 Book a call

    Next stage: confirm implementation kickoff, owners, and technical credential handover for week one.

    3.3 FAQ

    1) Is Zeus AI just ChatGPT with a new name?

    No. Zeus AI uses the latest OpenAI models, but the product is much more than a model wrapper.

    Zeus includes its own workflow logic, guardrails, integrations, and a custom knowledge layer (retrieval-augmented generation) so responses are grounded in your business context.

    In practice, it behaves more like a trained digital team member than a generic chatbot.

    2) Which models power Zeus AI?

    Zeus AI uses the latest OpenAI models and we continuously upgrade as better models become available.

    Our architecture is model-agnostic, so we can improve model performance over time without requiring you to rebuild your workflows.

    3) How do you test for accuracy, robustness, bias, and hallucinations?

    This is a core part of our system design.

    We run an automated testing framework that executes prompts thousands of times across realistic and edge-case scenarios. We then iterate prompts, flows, and guardrails until performance meets expected standards.

    Our approach includes:

    • High-volume automated testing: repeated prompt runs at scale to measure consistency and failure modes.
    • Repeat-until-pass methodology: prompts and flows are tested and refined until they perform as expected.
    • Robust guardrails: Zeus is intentionally constrained to business-relevant context and approved behaviours.
    • Grounded answers via knowledge retrieval: for business-specific questions (e.g. pricing, policies), Zeus checks your custom knowledge base.
    • "Don't know" behaviour: if the information is not available, Zeus is designed to say it does not know rather than inventing an answer.
    • Controlled scope: Zeus is built to feel human, but remain operationally disciplined and domain-limited.

    This is how we reduce hallucinations while keeping conversations natural and useful.

    4) Is Zeus AI secure and enterprise-ready from a data/compliance perspective?

    Yes. Zeus is built for enterprise use and we regularly support large corporate security and procurement reviews.

    We can cover security questionnaires and controls around customer data processing, storage, logging, access, and segmentation.

    Key controls include:

    • Data protection and encryption
    • Auditability and logging
    • Data segmentation between clients
    • Information security review support for enterprise onboarding

    5) Is Zeus AI GDPR compliant?

    Yes. Zeus AI supports GDPR compliance requirements, including consent handling, access, and erasure workflows.

    If required, we also support corporate information security processes to document how customer data is handled end-to-end.

    6) How does Zeus integrate with existing systems?

    Zeus supports both low-effort and advanced integration paths.

    Low-effort pilot (fast start):

    • Upload a CSV/spreadsheet of leads
    • Zeus runs outreach and qualification
    • Call/meeting requests are sent by email
    • No complex training or deep technical setup required

    Advanced integration (production):

    • API to ingest leads in real time (e.g. PPC, forms, third-party referrals)
    • Webhooks to push qualified leads and outcomes into CRMs/systems (e.g. Salesforce, Monday, proprietary CRM)
    • Live diary booking links for direct scheduling
    • Tailored lead-state logic (e.g. unsubscribe rules, pause/resume nurture, reactivation timing)

    7) What can Zeus actually do for the business?

    Zeus can proactively engage contacts, qualify intent, answer business questions using approved knowledge, nurture leads over time, and book interested prospects into calls.

    The goal is to increase speed-to-lead, reduce manual follow-up load, and convert more opportunities consistently.

    8) What support do we get after implementation?

    You get dedicated support, including onboarding and practical rollout guidance.

    For larger deployments, we provide structured account management and priority support paths so Zeus remains reliable as a core operating system.

    Testimonials