A twin beside
every role.
One digital twin for every person, designed around the way the organisation already works. Agents prepare. People decide.
Organisation-mapping figures supplied in the case-study brief. The client identity is withheld.
Your best people should drive growth,
not chase coordination.
Routine work crowds out selling
CRM updates, meeting preparation, pipeline reports and follow-ups pull account teams away from customers.
Hand-offs disappear into inboxes
Work moves across marketing, sales, pre-sales and product without one shared view of what is waiting.
Decisions wait on leaders
Questions reach the right person late, with too little context and no consistent way to track turnaround.
Goals lack an operating layer
KPIs exist, but ownership, dependencies, routines and evidence are difficult to see in one place.
A parallel agent organisation,
built around the human one.
The system mirrors roles and responsibilities. It does not replace them. A control layer sits between every agent proposal and every real-world action.
41 people across 10 teams
People retain ownership of customer relationships, decisions and final actions.
Named approvals and policy gates
Every proposal is checked, routed and held for the right person before anything moves.
One twin beside every role
Each twin works from its role card, KPIs, routines and authorised organisational context.
KPI sheet
Start with the scorecard the organisation already uses.
Role card
Define what each person owns and how success is reviewed.
Hand-off map
Name what moves between roles, who sends it and who receives it.
Success streams
Group work around shared goals, owners and contributors.
Agent playbooks
Turn recurring work into reviewable daily and weekly routines.
Less chasing.
More coordinated action.
A gap becomes visible
The twin finds stalled deals and a partner registration that needs attention.
A prioritised issue, not another dashboard alert.
Illustrative workflow using sample account activity. It demonstrates how work moves through the agent organisation, not a reported client result.
Agents propose.
People remain accountable.
The system is intentionally capped at L2 autonomy. It can draft, recommend and route. A person approves the decision and owns what happens next.
Draft first
Twins prepare work. People decide whether it moves.
Named owner
Every twin belongs to a role with a person accountable for its actions.
Two keys for higher risk
Customer-facing and sensitive work can require a second approval.
Grounded context
Twins work only from authorised organisational data and state what is missing.
Central stop control
Agent routines can be paused while people review the system.
Measured against
the organisation's own KPIs.
These are the measures built into the operating model. They describe how value is evaluated, not unverified performance gains.
Decision turnaround
How long priority decisions wait before a named owner resolves them.
Pipeline hygiene
Open opportunities with a current owner, next step and date.
Customer follow-through
Drafts reviewed and customer actions completed within the agreed rhythm.
Hand-offs on time
Cross-team work delivered when the receiving role needs it.
Time returned
Routine work no longer completed manually, validated by the people doing the job.
Approval quality
Approval rates, rejection reasons and policy flags used to improve the playbooks.
Assess
Map roles, KPIs and the highest-friction hand-offs.
Design
Agree the first agent pod, policies and approval model.
Build
Connect context and turn routines into governed playbooks.
Deploy
Start with a small pod at draft-only autonomy.
Success
Measure against the baseline, then scale what earns its place.
Make AI earn its place in every role.
Start with one team, one scorecard and one set of hand-offs. Build the first pod around work people can review and value they can measure.