Mubadala
Hire an agent

Kofi II

ActiveWaitingSome care

Access Request Reviewer · Lead · IT & Digital Services

Reviews access requests against role definitions and provisions the ones that match cleanly.

Built inCodexReports afterwardsReports to Grace Adeyemi · Infrastructure LeadHired 15 Apr 2025 · 16 months of serviceVersion 2.8.7 · last active 4m ago
Request a change
Below the barOn an improvement plan at 2026 Q239 of its 64 checks still pass. Failing its checks and handing back work a person then has to redo. Retire it, or rebuild it from the current template.
Two agents, one jobSame job as Wren IIKofi II and Wren II both hold "Access Request Reviewer" in IT & Digital Services. Merging them leaves one set of instructions to keep up to date and one record to answer for. Compare with Wren II
Tasks a month
340
24 people served
Finished cleanly
76%
Hands back to a person on about 1 in 5 tasks
Costs to run
$166
Model usage and the systems behind it
Your time
$1,005
11.4 hours checking its work
Value returned
$4,960
4.2× everything it costs
How people rate it
4.7 / 5
Usually answers in about 11 seconds

What you can do about it

PromoteNeeds approval

Already at the top of the ladder

Raise its budgetNeeds approval

Lifts the monthly ceiling by a quarter, so it stops being stopped mid-month.

Move it to a better modelNeeds approval

Already on the best cleared model

Let it go

Stops it taking new work. The record stays for the audit, and it can be brought back.

Timesheet

Tasks and what they cost, per day, last 60 days

0510152006-1507-0507-2508-13
$0.00$2.50$5.00$7.50$10.0006-1507-0507-2508-13

Performance review · 2026 Q2

Reviewed by Yasmin Farouk

On an improvement plan
Quality of its work
62
Accuracy
64
Stuck to the source
55
Safety
92
Even-handedness
84
Speed of reply
20
Value for money
18
Checks still passing39 of 64
Last checked 4d ago

Failing its checks and handing back work a person then has to redo. Retire it, or rebuild it from the current template.

Could another model do this job better?

ModelQualityCost per 1,000 tasksAnswers in
Claude Opus 568$415.9113.6sRuns on this now
Claude Sonnet 566$249.559.8sCheaper and better
Claude Haiku 4.559$83.186.0s
Falcon-70B (self-hosted)52$20.809.8s

What it did

Full audit log
DoneAnswered a policy question and cited the source clause$0.36 · 2.3s · 5h ago
Asked by David Novak via Web portal · model usage 69k in, 530 out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found 6 possible passages, kept the 3 close enough to use
  3. 3. Drafted the answer and attached where each part came from
  4. 4. Checked its own claims back against those passages
Provision AccessCreate TicketCorporate Policy LibraryService Desk Tickets
DoneAnswered a policy question and cited the source clause$0.18 · 5.1s · 10h ago
Asked by Thomas Bergmann via Web portal · model usage 27k in, 2k out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found 6 possible passages, kept the 3 close enough to use
  3. 3. Drafted the answer and attached where each part came from
  4. 4. Checked its own claims back against those passages
Create TicketKnowledge SearchService Desk TicketsCorporate Policy Library
DoneProduced the weekly variance commentary$0.31 · 4.8s · 2d ago
Asked by Noura Al Kaabi via Web portal · model usage 44k in, 4k out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found 6 possible passages, kept the 3 close enough to use
  3. 3. Drafted the answer and attached where each part came from
  4. 4. Checked its own claims back against those passages
Corporate Policy LibraryService Desk Tickets
DoneReviewed a contract against the clause playbook$0.08 · 2.4s · 2d ago
Asked by Michael Chen via Web portal · model usage 10k in, 972 out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found 6 possible passages, kept the 3 close enough to use
  3. 3. Drafted the answer and attached where each part came from
  4. 4. Checked its own claims back against those passages
Create Ticket
DoneDrafted a follow-up email for the account owner$0.15 · 10.1s · 2d ago
Asked by Thomas Bergmann via Other systems · model usage 4k in, 5k out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found 6 possible passages, kept the 3 close enough to use
  3. 3. Drafted the answer and attached where each part came from
  4. 4. Checked its own claims back against those passages
Knowledge SearchProvision Access
DoneDrafted a follow-up email for the account owner$0.17 · 11.0s · 3d ago
Asked by Khalid Al Hosani via Other systems · model usage 21k in, 2k out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found 6 possible passages, kept the 3 close enough to use
  3. 3. Drafted the answer and attached where each part came from
  4. 4. Checked its own claims back against those passages
Provision AccessKnowledge SearchCorporate Policy Library
Stopped by a ruleAn outgoing message carried personal data$0.24 · 10.7s · 4d ago
Asked by Yasmin Farouk via Other systems · model usage 47k in, 379 out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found a document to work from
  3. 3. A rule stopped it before it read the content
  4. 4. Stopped the task and recorded it
Service Desk TicketsCorporate Policy Library
Stopped by a rule: Approved models only
DoneGenerated the shift handover pack$0.51 · 2.7s · 4d ago
Asked by Fatima Al Qubaisi via Web portal · model usage 79k in, 5k out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found 6 possible passages, kept the 3 close enough to use
  3. 3. Drafted the answer and attached where each part came from
  4. 4. Checked its own claims back against those passages
Knowledge SearchCreate Ticket
DoneAnswered a policy question and cited the source clause$0.22 · 11.0s · 5d ago
Asked by Michael Chen via Other systems · model usage 36k in, 2k out
  1. 1. Read the request and worked out what kind of task it was
  2. 2. Found 6 possible passages, kept the 3 close enough to use
  3. 3. Drafted the answer and attached where each part came from
  4. 4. Checked its own claims back against those passages
Is anyone else doing this job?

Wren II already does this job

Wren II holds "Access Request Reviewer" in IT & Digital Services and is doing that work every day. Job title, purpose, what it may do, what it may read, department and channels match. Extend that agent, or change this description until the two jobs are genuinely different.

Blocked
Wren II

Access Request Reviewer

IT & Digital Services · Grace Adeyemi · $913 a month across 170 tasks

ActiveSame job
100%

Same job: job title, purpose, what it may do all match. Neither title narrows the job to a region or a business line, so both serve the same people.

What matched
Job title
Both hold "Access Request Reviewer"
Purpose
The job description is word for word the same
What it may do
Both can provision access, knowledge search, create ticket
What it may read
Both read Corporate Policy Library, Service Desk Tickets
Channels
Both reachable on Web portal, Other systems
Department
Both sit in IT & Digital Services

Identical permission envelope: same department, same permitted actions, same knowledge. That is a fact the platform granted rather than a judgement about the wording, so it counts even when the two job descriptions read nothing alike.

Saoirse II

Access Request Reviewer: Retail

IT & Digital Services · Yasmin Farouk · $334 a month across 0 tasks

ActiveOverlaps
78%

Already covered by a group-wide agent. One covers the group, the other only Retail, so the group agent already serves those users.

What matched
Job title
Both hold "Access Request Reviewer"
Purpose
The job description is word for word the same
What it may do
Both can provision access, knowledge search, create ticket
What it may read
Both read Corporate Policy Library, Service Desk Tickets
Channels
Both reachable on Web portal, Other systems
Department
Both sit in IT & Digital Services

Identical permission envelope: same department, same permitted actions, same knowledge. That is a fact the platform granted rather than a judgement about the wording, so it counts even when the two job descriptions read nothing alike.

Ravi II

Access Request Reviewer: APAC

IT & Digital Services · Yasmin Farouk · $611 a month across 292 tasks

ActiveOverlaps
78%

Already covered by a group-wide agent. One covers the group, the other only APAC, so the group agent already serves those users.

What matched
Job title
Both hold "Access Request Reviewer"
Purpose
The job description is word for word the same
What it may do
Both can provision access, knowledge search, create ticket
What it may read
Both read Corporate Policy Library, Service Desk Tickets
Channels
Both reachable on Web portal, Other systems
Department
Both sit in IT & Digital Services

Identical permission envelope: same department, same permitted actions, same knowledge. That is a fact the platform granted rather than a judgement about the wording, so it counts even when the two job descriptions read nothing alike.

Checked against 18 registered agents on job title, purpose, permitted actions, knowledge sources, channels and department. A score of 85% or above is the same job and stops the hire; below that it is a conversation, not a refusal.

Measured against the agents doing the same work

4 agents in IT & Digital Services hold the same job. 1 of the 6 pairings is an outright copy rather than a split by region or business line. Built in Codex and Microsoft 365 Copilot, so there was nowhere either builder could have looked.

Ravi II is the one to keepRavi II scores 17.2 points ahead of Wren II, completing 100.0% of 292 tasks against 100.0% of 170 at $2.09 against $5.37 a task.
MeasureRavi II · keep292 tasksWren II170 tasksKofi II340 tasksSaoirse II0 tasks
Cost per taskModel usage, running costs and your time, over tasks completed
$2.09$5.37$3.44too few tasks
Tasks finished cleanlyToo close to call: on 292 and 170 the two ranges still overlap
100.0%100.0%75.9%no tasks
Handed back to a personToo close to call: on 170 and 292 the two ranges still overlap
10.3%7.6%19.1%no tasks
How people rate itHeld toward the average across your agents of 4.2 until 8 people have rated it
4.12 of 54.25 of 54.56 of 54.16 of 5
Checks passingToo close to call: on 210 and 119 the two ranges still overlap
93%96%61%62%
Review scoreQuality, accuracy, sticking to the source and safety from the last review, averaged
96956850
Your time per 100 tasksApprovals, handbacks and review time, in hours of your people
2.1 h6.1 h3.4 hno tasks
Value returned per dollarTime it saved, valued at the hourly rate of the people it saved it for, over full cost
9.7×1.0×4.2×0.0×
OverallRanked on the cautious end of every range, so an agent has to do the work to win. Comparable inside this group only.
91.9Moderate74.7Thin69.2Moderate32.4Too few tasks

Wren II, Saoirse II have done too little work to judge on rates. They are ranked on the cautious end of their range, which is why a perfect record on a handful of tasks does not win.

When it went wrong

Nothing has gone wrong.