Hamblett Consultancy

AI OPERATIONS ASSISTANT

Meet Lucy: AI That Works Inside the Business

Lucy is Hamblett Consultancy’s AI operations assistant. She connects approved systems, follows defined processes, checks her work and brings people the decisions that actually need them.

This is not a website chatbot or a generic AI subscription. It is a practical example of how AI can become part of a business operating model without being given uncontrolled access.

Live inside HC operations Human accountable Approval-gated Built around real systems
Lucy, Hamblett Consultancy AI operations assistant
Lucy works as part of the Hamblett Consultancy team.

THE SHORT ANSWER

What is an AI operations assistant?

An AI operations assistant is a governed digital team member that can understand business information, follow repeatable workflows, use approved tools and report or escalate the outcome.

Lucy supports Hamblett Consultancy across service delivery, marketing, administration, research, documentation and communications. Every capability is connected separately, tested against real work and limited by defined permissions.

HOW THE MODEL WORKS

One operating loop, many business workflows

Useful business AI does more than generate text. It needs a trigger, trusted context, controlled action, independent checking and a clear route back to a person.

The governed operating loop used across Lucy’s workflows.
01
TriggerEmail · alert · call

A new request or signal enters an approved workflow.

  • Capture the request
  • Record key details
  • Acknowledge receipt
02
UnderstandContext · rules

Lucy gathers the right context and applies the business rules.

  • Check history
  • Read approved context
  • Set priority
03
ActDraft · classify

The next permitted action is prepared or completed.

  • Follow the workflow
  • Record the action
  • Pause at approvals
04
VerifyEvidence · outcome

Lucy checks that the intended result actually happened.

  • Test the outcome
  • Attach evidence
  • Spot exceptions
05
ReportComplete · escalate

The outcome is reported clearly, with a human route when needed.

  • Update the record
  • Notify the owner
  • Escalate if unresolved
ESCALATEif unresolved

Outcome confirmedEvidence retained. People stay accountable.

Human control rail People set the rules Approvals stop risky actions Exceptions return to the team Evidence remains reviewable

SERVICE OPERATIONS

Turning service noise into an actionable queue

Lucy can read approved service information, separate routine events from genuine exceptions, watch deadlines and prepare clear updates. The aim is not to hide support work behind AI; it is to help the team see the right work sooner.

  • Ticket and alert triage
  • SLA and neglected-case monitoring
  • Repeat-fault and trend detection
  • Client update preparation
  • Escalation when judgement is required

HC position: core systems are connected; expanded helpdesk operating workflows are being introduced in controlled stages.

Service operations flow
Tickets Alerts SLAs
TRIAGE LAYER Classify · correlate · check urgency
ROUTINERecord & monitorKnown event · within policy
WATCHTrack the deadlineSLA risk · recurring signal
ESCALATEBring in judgementImpact · ambiguity · exception
One prioritised service queueClear owner · next action · evidence

SALES & MARKETING

Consistent growth work without losing the brand

Lucy researches opportunities, supports prospecting, creates and validates content, manages publishing workflows and watches the follow-up queue. The brand remains the source of trust: Lucy operates as part of Hamblett Consultancy, not as a disconnected bot.

  • Prospect research and qualification
  • Campaign preparation and controlled follow-up
  • Website, SEO and answer-engine optimisation
  • Social planning, publishing and verification
  • CRM hygiene and lead progression

For another business, the same model can keep sales activity moving while enforcing contact rules, exclusions, approval thresholds and tone of voice.

Lead-to-learning flywheel

ADMINISTRATION & COMMERCIAL CONTROL

Repetitive checks handled; exceptions made visible

Licensing, renewals, reminders, records and reports often fail through delay rather than complexity. Lucy can compare approved source systems, maintain mappings and surface only the cases that do not match confidently.

  • Licence and renewal follow-up
  • Billing reminder workflows
  • Customer and contact matching
  • Scheduled reports and daily briefs
  • Exception queues instead of spreadsheet archaeology

Financial treatment, ambiguous matches and unusual transactions remain human decisions. Automation should make accountability easier, not blur it.

Exception-first administration
CRMCustomers
FINANCEBilling
LICENSINGRenewals
EMAILResponses
CONTROL LAYERMatch · check
classify
Confidence scored

CONFIDENTProcess or report

UNRESOLVEDAsk a person

KNOWLEDGE & DECISION SUPPORT

From scattered information to an evidence-backed answer

Lucy can search approved knowledge, compare sources, explain what is known and distinguish current evidence from an assumption. Useful findings can then become documentation, a brief, a process note or a decision for the team.

  • Technical and commercial research
  • SharePoint and document discovery
  • Source-backed summaries and briefs
  • Living documentation and stale-content checks
  • Memory that records decisions, not secrets
Evidence-to-decision flow
SOURCES

KnowledgeApproved documents

Live systemsCurrent operational data

Public evidenceVerifiable sources

RETRIEVE→COMPARE→CITE
EVIDENCE LEDGERKnownCurrentTraceable
ANSWERDOCUMENTDECISION

VOICE & COMMUNICATIONS

A telephone conversation, not a novelty recording

Hamblett Consultancy has proved a live, low-latency voice path that connects its existing Yay.com phone platform to OpenAI Realtime through a protected Asterisk and audio bridge. Lucy identifies herself, responds naturally and can be interrupted.

The current proof is deliberately isolated from business tools and long-term memory. That separation lets the conversation layer be tested before read-only context, call summaries and approval-gated actions are added.

See our business VoIP service
Live voice architecture
PUBLIC TELEPHONE NETWORK

CallerNatural speech

BUSINESS TELEPHONY
YAY

Yay.comNumber · routing · caller ID

PROTECTED HC VOICE BRIDGE
AX

AsteriskSecure call control

AB

Audio bridgeStreaming · guardrails

REALTIME CONVERSATION
AI

OpenAI RealtimeMarin · interruption aware

Identifies as AI No business tools Call limits
Resilient fallbackLocal Whisper recognition → governed reasoning → Sonia voice

ON-SITE SYSTEMS & PRIVATE ACCESS

Reach approved servers and NAS information without exposing them publicly

Some useful business information never lives in Microsoft 365 or a cloud application. It may sit on an office server, a NAS box, an internal dashboard or a system that is deliberately unavailable from the public internet.

Where a business approves it, Tailscale can create a private encrypted path between Lucy’s controlled environment and specific on-site devices. Lucy can then work with the folders, services or monitoring data she has been explicitly permitted to use.

  • Read approved documents and knowledge from a NAS
  • Collect server health, logs or inventory information
  • Check backup status and surface exceptions
  • Use private dashboards or internal APIs
  • Keep public inbound ports closed

Important: Tailscale provides the private network route. Device identity, access policies and the server or NAS account still limit what Lucy can read, change or execute.

See our managed IT support service
Private access to approved on-site systems
CONTROLLED AI ENVIRONMENT

LucyApproved workflow only

YOUR PRIVATE BUSINESS NETWORK

On-site serverLogs · services · internal data

NAS storageApproved folders · documents

Private serviceDashboard · monitoring · API

No public inbound port Named devices only Least-privilege account Human approval for change

Access boundaryNetwork access never overrides the permissions on the server, NAS or application.

WHAT THIS COULD DO FOR YOUR BUSINESS

Build around your systems, people and commercial priorities

The useful question is not “Where can we add AI?” It is “Which repeated business problem is worth solving, and what controls must surround it?”

01

Service desk assistant

Triage tickets, monitor SLAs, prepare updates and highlight recurring faults.

02

Sales operations assistant

Qualify enquiries, maintain follow-up, prepare proposals and keep the CRM accurate.

03

Client-success assistant

Combine service, licensing, security and billing signals into useful account reviews.

04

Knowledge assistant

Find approved information, answer internal questions and keep documentation current.

05

Voice and reception assistant

Identify callers, capture intent, route enquiries and create structured follow-up.

06

Back-office workflow assistant

Handle repeatable checks, reminders, exceptions and reports across connected systems.

A staged route from one business problem to governed autonomy.
  1. 01
    FRAME THE VALUEDiscover

    Choose one valuable workflow

  2. 02
    LIMIT THE SURFACEConnect

    Read-only, least-privilege access

  3. 03
    PROVE THE JUDGEMENTShadow

    Compare decisions with your team

  4. 04
    CONTROL THE ACTIONApprove

    Permit specific bounded actions

  5. 05
    EXPAND WITH EVIDENCEImprove

    Measure outcomes and expand carefully

GOVERNANCE FIRST

The controls matter as much as the capability

Business AI should earn authority. Hamblett Consultancy uses a staged model so access and autonomy increase only when a workflow has been proved.

Read our AI Privacy Policy
AI authority increases only after each stage is proved.
MORE AUTHORITY
  1. 1
    ReadInspect approved information
    LOW RISK
  2. 2
    RecommendWork in shadow mode
    PROVE
  3. 3
    Supervised actionAct with approval
    APPROVE
  4. 4
    Bounded autonomyAct inside proven rules
    MEASURED
Human accountability continues at every levelExceptions always return to the team

COMMON QUESTIONS

Lucy AI and AI operations assistants

What is Lucy AI?

Lucy is Hamblett Consultancy’s AI operations assistant. She works through controlled workflows that connect approved business systems, apply defined rules, verify outcomes and escalate exceptions to people.

What does Lucy do for Hamblett Consultancy?

Lucy supports marketing and website operations, research, prospecting and campaign workflows, administration, service monitoring, documentation, reporting and a live voice proof of concept. Each workflow has its own permissions and safeguards.

Can an AI operations assistant work in another business?

Yes. The approach can be adapted to a business’s existing helpdesk, CRM, Microsoft 365, telephony, website, finance, licensing and knowledge systems. It should start read-only, prove its decisions in shadow mode and gain bounded authority only when the workflow is reliable.

Does Lucy replace employees?

No. Lucy is designed to remove repetitive handling, maintain follow-up and surface useful decisions. People remain accountable for sensitive, unusual, contractual, financial and high-impact work.

How is business data protected?

Access is limited by task, credentials are protected, external actions are controlled, important results are verified and sensitive or ambiguous decisions are escalated. Hamblett Consultancy’s AI Privacy Policy explains the wider data and accountability approach.

Can Lucy speak on the telephone?

Yes. Hamblett Consultancy has proved a low-latency telephone workflow using Yay.com, Asterisk and OpenAI Realtime. Lucy identifies herself clearly as an AI assistant and the proof remains restricted while memory and business-system access are developed separately.

Can Lucy securely access an on-site server or NAS?

Yes, where the business approves it. Tailscale can provide a private encrypted route to named servers, NAS devices or internal services without publishing them directly to the internet. Tailscale access policies, device identity and the server or NAS permissions still determine exactly what Lucy can read or do.

START WITH A REAL BUSINESS PROBLEM

What would you stop doing manually?

We can map the workflow, systems, decisions and safeguards before anyone starts buying shiny AI licences.