Skip to content
HeuriTel
Sign in
Delivery

A working capability in ninety days, handed to your own team.

A telecom-specialist AI team that delivers a working capability in 90 days and hands it over. Bought by CTIO and Chief Digital Officer; measured on capability delivered and handover accepted by your own team.

The plan, the roles, the pods and the handover below are the operating model’s own lines.

Scope a controlled pilot Read the offer
The ninety days

14 phases, each with what ends it and who answers for it.

Weeks as the live runbook states them. A phase ends on its exit evidence, not on a date; each carries the rule that stops it.

  1. 01 Commercial qualification

    Confirm exact product, buyer, problem, value unit and authority boundary

    Ends on
    Signed discovery brief; no unverifiable outcome promise
    Answers for it
    Executive sponsor
    Stops it
    Stop if buyer/authority/value unit is absent

  2. 02 Mobilisation

    Appoint client and HeuriTel owners; RACI; cadence; access plan

    Ends on
    Mobilisation pack approved
    Answers for it
    Engagement lead
    Stops it
    Escalate unowned dependencies

  3. 03 Use-case contract

    Define trigger, population, action alternatives, stop conditions and KPI denominator

    Ends on
    Use-case contract signed
    Answers for it
    Product owner
    Stops it
    Do not start build with ambiguous action authority

  4. 04 Data and privacy assessment

    Attribute map, lawful basis/consent, minimum data, quality and retention

    Ends on
    Data readiness and privacy gate
    Answers for it
    Data owner + privacy
    Stops it
    Use synthetic data until approved

  5. 05 Architecture and integration

    Map systems of record, APIs, security, idempotency and failure paths

    Ends on
    Architecture/security approval
    Answers for it
    Solution architect
    Stops it
    No direct production access before approval

  6. 06 Baseline and control design

    Declare denominator, baseline, control/holdout, value formula and guardrails

    Ends on
    Measurement plan approved by finance/business
    Answers for it
    Analytics lead
    Stops it
    No uplift claim without design

  7. 07 Synthetic thin slice

    Run end-to-end with synthetic attributes and mocked responses

    Ends on
    Demo covers success, refusal and failure
    Answers for it
    Engineering lead
    Stops it
    Fix unsafe/default-open paths

  8. 08 Data/logic build

    Build features, rules/model, reason codes and monitoring

    Ends on
    Offline validation and data quality passed
    Answers for it
    Data/ML lead
    Stops it
    Reject leakage, unstable labels or prohibited attributes

  9. 09 Integration build

    Decision API, source-system calls, channel and outcome capture

    Ends on
    Contract/integration tests pass
    Answers for it
    Integration lead
    Stops it
    Fail closed on consent/eligibility/price

  10. 10 Quality and security

    Functional, negative, load, privacy, security, model and rollback tests

    Ends on
    Release evidence approved
    Answers for it
    QA + security
    Stops it
    Critical defects block pilot

  11. 11 Shadow run

    Score/decide without customer or network action; compare with current process

    Ends on
    Shadow stability and exception review
    Answers for it
    Product owner
    Stops it
    Tune or rescope; do not silently widen authority

  12. 12 Controlled pilot

    Start small, preserve control, apply spend/exposure caps and daily review

    Ends on
    Pilot gate metrics within guardrails
    Answers for it
    Business authority
    Stops it
    Pause on harm, loss, complaints or system errors

  13. 13 Operational acceptance

    SLO, runbook, alerts, support, roles, audit retrieval and reconciliation

    Ends on
    Operational acceptance signed
    Answers for it
    Service owner
    Stops it
    No scale without on-call ownership

  14. 14 Handover and training

    Transfer code/config, playbook, monitoring, evidence and backlog

    Ends on
    Handover checklist and training complete
    Answers for it
    Delivery lead
    Stops it
    Document residual risks

After the pod

  • Scale decision (post-pilot) — Review incremental outcome, guardrails, reliability and unit economics. Ends on scale, revise or stop decision recorded; steering committee answers for it. A failed hypothesis is an acceptable outcome.
  • Live operations (ongoing) — Monitor data/model/integration/business performance; approve changes; incident and value reviews. Ends on monthly service and value pack; service manager answers for it. Rollback/retrain/reapprove when thresholds breach.
The pod and your own roles

15 roles, and who is accountable in each phase.

A — accountable, R — responsible, C — consulted, I — informed, as the operating model’s RACI states them. Your own owners keep the authority; the pod delivers within the agreed statement of work.

RoleSideDiscoveryBuildPilot gateLive runOwns
Executive SponsorYoursAACIOwns business outcome, funding and escalation
Client Product OwnerYoursA/RA/RA/RA/RDefines scope, priority and acceptance
Authority/System OwnerYoursCA/RA/RA/ROwns charging, credit, network, agency or transaction authority
HeuriTel Product LeadThe podRRCCMaintains product contract and value logic
Telecom Solution ArchitectThe podRRCCMaps systems, APIs, security and failure paths
Telecom AI CoordinatorThe podRRRRCoordinates cross-location delivery, dependencies and evidence
Data EngineerThe podRRCRBuilds approved data products and quality controls
ML/Decision EngineerThe podCRCRBuilds and monitors models/rules/reason codes
Integration DeveloperThe podCRRRImplements source-system and execution interfaces
QA Automation EngineerThe podCRRROwns functional, negative, regression and performance evidence
DevSecOps/SREThe podCRRRCI/CD, environments, observability, incidents and rollback
Privacy/Compliance/Model RiskYoursCCA/RCApproves lawful use, controls, validation and monitoring
Finance/Measurement AnalystYoursRCCRApproves denominator, value formula and control readout
Local Market SMEEitherCRRRValidates language, process, product and market rules
Service ManagerYoursICCA/ROwns SLO, support, change and monthly service review
The pods

10 pods, one commercial model each.

The ninety-day pod leads; the rest are shaped to a use case, a market wave or the run after the pilot. Each states its scope, its team, where it works from, what it hands over, what it needs from you and how it is priced.

POD-01 · 12 weeks

90-Day Telecom AI Deployment Pod

One bounded use case; one market; synthetic demo to controlled pilot.

HT-0567 in the catalogue
Team
0.5 engagement lead; 1 telecom product lead; 1 architect; 1 data engineer; 1 ML/AI engineer; 1-2 integration developers; 1 QA; 0.5 DevSecOps; 1 coordinator — 6.5-8.5 FTE.
Works from
Remote/nearshore with planned on-site discovery and pilot weeks.
Hands over
Accepted pilot, code/configuration, tests, runbook, evidence pack, training.
Needs from you
Client sponsor, product owner, data/system owners, security access.
Commercial model
Fixed-duration pod fee + travel.

  1. 120-Day Telecom AI Deployment Pod16 weeks · 8.5-10 FTE

    Complex use case or two related use cases; one market.

    Hands over
    Pilot plus hardening, operating dashboard and handover.
    Needs from you
    Same as POD-01 plus production-like environment.
    Commercial model
    Fixed-duration or milestone price.

  2. CAMARA API Delivery Pod12-16 weeks · 7 FTE

    First 1-3 APIs, portal/sandbox, conformance and partner pilot.

    Hands over
    Conformant API, sandbox, metering, operational runbook, partner pilot.
    Needs from you
    Existing network capability and API/security owner.
    Commercial model
    Setup + per-API fee; optional managed operations.

  3. Contact Centre AI Deployment Pod12-16 weeks · 6.5-8 FTE

    Three to five bounded L0/L1 intents in voice/chat.

    Hands over
    Tested intents, handoff, QA dashboard, knowledge process.
    Needs from you
    CCaaS access, approved knowledge, human transfer queue.
    Commercial model
    Implementation + usage subscription.

  4. CVM Managed Execution Pod16 weeks initial · 6.5 FTE

    One revenue or retention programme operated with client CVM team.

    Hands over
    Live campaigns, control design, outcome ledger, playbook.
    Needs from you
    Approved offers, channels, stable data and finance formula.
    Commercial model
    Monthly managed service + product subscription.

  5. Credit Propensity Delivery Pod14-18 weeks · 7 FTE

    Shadow model and capped lender-approved pilot.

    Hands over
    Approved feature list, validation, reason codes, pilot and monitoring.
    Needs from you
    Regulated lender authority, consent, labels, appeals.
    Commercial model
    Build/validation fee + per-decision/monitoring.

  6. Network AI Delivery Pod12-16 weeks · 7 FTE

    One planning or operations decision-support scenario.

    Hands over
    Shadow/pilot recommendation workflow and KPI verification.
    Needs from you
    Telemetry, topology, OSS owner, change authority.
    Commercial model
    Implementation + annual licence.

  7. Government Data Use-Case Pod12-16 weeks after approvals · 7 FTE

    One lawful aggregated analytics or citizen-service workflow.

    Hands over
    Purpose gate, privacy design, pilot, audit and impact report.
    Needs from you
    Legal basis, data agreement, agency action owner.
    Commercial model
    Government project + managed service.

  8. Multi-Country Rollout Pod16 weeks per wave · 7.5-10 FTE

    Reuse proven product across 2-4 markets.

    Hands over
    Country packs, regression evidence, cutover and local training.
    Needs from you
    Reusable interfaces, group governance, market owners.
    Commercial model
    Wave price + local travel/compliance.

  9. Telecom AI Managed Operations PodOngoing after pilot · 4.5 FTE

    Operate models, decisions, incidents and monthly value reviews.

    Hands over
    SLO reports, incidents, change/retraining evidence, value reviews.
    Needs from you
    Production acceptance and support model.
    Commercial model
    Monthly managed-service fee by scope/SLA.
Build, operate, transfer

The work ends in your hands, or in a managed run you chose.

Client product, security and acceptance owners retain authority; the pod delivers within the agreed SOW.

  1. Outcome backlog
  2. Fortnightly increments
  3. Architecture/data/security gates
  4. Demonstration
  5. Acceptance
  6. Handover or managed run

Conditions that stop the work

  • No access
  • No owner
  • Unapproved data
  • Target API absent
  • Security onboarding delay
  • Acceptance ambiguity
  • Local labour/travel restriction

Handover requirements

  • SOW
  • RACI
  • Backlog
  • Architecture
  • Code/test results
  • Demos
  • Acceptance
  • Runbook
  • Training and handover record

Pricing models

  • Fixed-duration pod fee
  • Time and materials with capacity cap
  • Milestone-based fixed price
  • Travel/local employment separately priced

A bounded, accepted pilot/increment plus reusable delivery assets and trained client owners.

Computed on the sample

A plan of 6 milestones, and what still blocks its handover.

The Team-as-a-Service engine run on its sample (HeuriTel default synthetic sample · Kenya): every milestone with its acceptance gate and who accepts it, and the list the operator’s owner would still need to see. The same computation runs on the offer page and in the workspace.

MilestoneWeekDeliverableAcceptance gateAcceptsState
M12Data contract agreed for the CRM extractExtract validated against the contract with zero structural refusalsCVM directoraccepted
M24Decisioning running on the sandbox extractTen decisions reviewed and agreed by the CVM teamCVM directoraccepted
M36Reconciliation of one closed monthFinance confirms at least one exceptionRevenue assurance managerin progress
M49Exception workflow with maker-checkerTwo exceptions closed through the workflowRevenue assurance managerplanned
M511Runbook and monitoringOperator engineers run one full cycle unaidedHead of data platformplanned
M612HandoverEvery scope item accepted with evidence; access revokedCTIOplanned
Before handover can be accepted — 7 items
  • Read access to the CRM and billing extracts in a non-production environment (partial, Head of data platform)
  • Security review of the pod's access and tooling signed (missing, CISO)
  • M2 Decisioning running on the sandbox extract — accepted, evidence not filed
  • M3 Reconciliation of one closed month — in_progress
  • M4 Exception workflow with maker-checker — planned
  • M5 Runbook and monitoring — planned
  • access revoked on every scope item

The offer, with its contract and pilot prerequisites The delivery stream in the catalogue