The console.

Sample data. Every figure is fictional.

Every page is behind the sidebar. Open a system from a star, a row or a card, and come back from the breadcrumb.

Overview

This weekThis monthThis quarter
Updated 12 s agoExport PDF

2 critical1 warning3 healthy6 systems, 18 agents

Success rate
98.4%▼ 0.7%
This week98.4%
Last week99.1%
Runs
57,580▲ 3.2%
This week57,580
Last week55,790
Spend
£9,583▲ 9%
Month to date£9,583
Last month£8,760
Compliance score
85▲ 3
This month85
Last month82

Estate

AI SDR agents4 agents · 8,400 runsSupply chain agents4 agents · 17,940 runsCustomer support agents3 agents · 24,600 runsFinance ops agents2 agents · 5,200 runsHiring agents2 agents · 260 runsMarketing agents3 agents · 1,180 runs
  • Bigger stars hold more agents
  • Brighter stars ran more this week
  • Critical
  • Warning

Priority actions

3 of 7
  • Supply chain agentsReliability

    Turn retry on for the OpenAI step

    1 in 25 runs fails there. 3 of 4 failed calls succeeded on retry.

    1 in 25 runs recovered

    Open
  • AI SDR agentsEU AI Act 14

    Add an approval step before outreach sends

    Lead details reach OpenAI with no human check.

    Closes 1 finding

    Open
  • Customer support agentsCost

    Move reply drafting to Gemini 3.8 Flash

    38 of 40 sampled replies matched GPT-6 Astra. 13 times cheaper per token.

    Save £1,140 a month, £13,700 a year

    Open

Explainability

41 explained events this week38 with a root cause3 with no reason recorded

  • Supply chain agents · Thu 18, 14:02

    Call OpenAI failed on 13 runs

    Because the provider returned rate-limit errors for 40 minutes. Not a prompt change. Inferred

    Trace
  • Hiring agents · Wed 24, 01:02

    Score the CV sent candidate details to OpenAI

    Because no reason recorded. The agent does not state why the scoring step chose the external model. Not recorded

  • Finance ops agents · Tue 23, 09:14

    Invoice matching held line 12 instead of posting

    Because the supplier name on the purchase order did not match the invoice, and the agent is told never to guess on a name. Agent stated

Failure causes

921 failed runs
  • Rate limited by provider410
  • Provider timeout240
  • Malformed reply130
  • Empty upstream response78
  • Prompt injection detected1
  • Unclassified62
  • Prompt injection: an input carried instructions aimed at the model. Flagged as a security cause. Nothing acted on it.
  • Unclassified: the error text matched no known cause. 62 to review.

Finance

  • GPT-6 Astra£4,120
  • Claude Opus 5£2,980
  • Gemini 3.8 Flash£1,640
  • Claude Fable 5.1£843
  • DeepSeek V4.1 Flash · self-hosted£0
  • Qwen3.8 27B · self-hosted£0

Projected £11,979 by 30 September, ▲ 37% against last month. Three cost recommendations would save £2,090 a month.

Compliance

85
of 100
  • 31 passing
  • 3 failing
  • 1 warning
  • ISO 42001 · UK GDPR · EU AI Act
  • Supply chain agents · API key written into an agentISO 42001
  • Hiring agents · Candidate data to an external model, no human checkEU AI Act 14

Security

  • 2 credentials at risk1 written into a node, 1 shared by 3 systems
  • 4 connections leave your cloudAll to OpenAI · 1 carries an open finding
  • 1 instruction aimed at the readerMarketing agents · nothing acted on it
  • Redaction on every stored payload0 raw payloads in the database

Systems

6 systems
SystemStatusAgentsRuns this weekSuccess rateAverage durationSpend, month to dateOpen issues