AccAnalysisAccAnalysis
Software & SaaS

Twenty agents, each doing one job properly, on one dashboard

A platform where repetitive knowledge work — email triage, CV screening, meeting notes, security scanning, outreach — is handled by a specific agent rather than by a general-purpose assistant that does everything approximately.

At a glance
Status
Delivered
Region
Global
Engagement model
Joint lab

Status: Delivered. Described from our own delivery records. Client identity, brand and commercial terms are withheld. Figures, where given, cover the period stated and nothing beyond it.

The organisation

Context

A software-as-a-service platform addressing a pattern common to almost every organisation: a long tail of repetitive tasks that individually do not justify a dedicated tool, and collectively consume a great deal of time.

The brief

The problem

  • Repetitive knowledge work — triaging email, scoring applications, writing up meetings, running routine security scans, handling outreach — was being done manually, in volume.
  • Each of those tasks is too small to justify buying a product for it, so they stay manual indefinitely.
  • General-purpose assistants handle each of them adequately and none of them well, because a good outcome depends on task-specific structure.
  • Running several point tools means several interfaces, several configurations and no single view.
The work

What we built

Twenty task-specific agents

Each built around one job — email triage, CV scoring, meeting transcription and notes, cybersecurity scanning, professional-network outreach, and others — rather than one agent asked to do everything.

Scheduled autonomous operation

So an agent runs on its cadence rather than waiting to be invoked.

A unified dashboard

Across all agents, so configuration, output and status sit in one place.

Per-agent pricing

So an organisation adopts the two agents that matter to it rather than a suite.

Method

How we delivered it

  1. 1

    Choose tasks by volume and rule-clarity

    The two properties that determine whether automation is genuinely worth it.

  2. 2

    Build a shared platform layer

    For scheduling, credentials, logging and the dashboard, so each agent is a specialisation rather than a rebuild.

  3. 3

    Build agents individually

    Each with its own evaluation set and quality bar.

  4. 4

    Instrument every action

    So what an agent did and why is inspectable.

  5. 5

    Release incrementally

    Agent by agent.

Sequence

How it was phased

PhaseDurationWhat happens
1Task selection
2 wks

Candidate tasks ranked by volume and rule clarity

2Platform layer
5–7 wks

Scheduling, credentials, logging, dashboard

3Agent waves
ongoing

Agents built and released individually, each with evaluations

4Instrumentation
2–3 wks

Action logging, output inspection, error handling

5Release
ongoing

Agent by agent

Indicative phasing for work of this shape. Actual duration varies with data quality, access and decision speed.

Hand-over

What the client keeps

  • The platform and agent source
  • Per-agent evaluation sets
  • Action logs
  • Scheduling and credential configuration
  • The dashboard
Stack
PythonLLM agentsScheduled orchestrationWeb dashboardAPI integrations
Keep reading

Related

Related case studies

Tell us what your current system can't do, and we'll tell you what it would take to change that.