9.1 Cap. Except as provided in Section 9.3, neither party's aggregate liability under this Agreement, including claims arising under Section 7 (Data Security), shall exceed the fees paid or payable in the twelve (12) six (6) months preceding the claim.

9.2 Exclusions. These limits do not apply to indemnification under Section 11, or to gross negligence.

9.3 Carve-outs. Claims arising under Section 7 (Data Security) are subject to the separate cap set out in the Term Sheet, and amounts payable under Section 11 shall not be counted against the cap in Section 9.1.

9.4 Survival. The limitations and exclusions in this Section 9 survive termination or expiry of this Agreement and apply to all claims, whether in contract, tort or otherwise, regardless of the form of action.

9.5 Aggregate application. Where more than one claim arises from a single course of conduct, those claims shall be treated as one claim for the purpose of the cap in Section 9.1, and the cap shall apply to the aggregate of all amounts payable in respect of them.

9.6 Insurance. Each party shall maintain cyber liability and professional indemnity cover of not less than the amount set out in Schedule 4 for the term of this Agreement and for two (2) years following its expiry, and shall provide evidence of that cover on written request.

9.7 Allocation of risk. The parties acknowledge that the limitations in this Section 9 reflect the allocation of risk agreed between them and are reflected in the fees, and that each party would not have entered into this Agreement without them.

✦reading §11 Indemnification…✦cross-checking the term sheet…

✦LUMENSHigh

The cap now covers data-breach claims.

“including claims arising under Section 7 (Data Security)”

The term sheet carves data-breach claims out of the cap.

Term sheet 3 · opened

2024 MSA 9.1 · unopened2024 MSA 9.1 · opened

Unlocks after both sourcesBoth sources opened

Lumens AI

An agent does the retrieval. A lawyer does the judgment. The product is the handoff.

Independent AI product · working build

AI REVIEW SYSTEM

FieldWake

Voice-first AI for field technicians whose hands are full, forty feet up a pole.

Independent AI product · in progress

VOICE

Approvals

This week: 7 closed by the agent, 4 needed a person.

Needs a person 4Agent working 2Closed 7Fixture runs 80

RequestSupplierAmountStep nameRequester

Attached quote from Slack: please buy 20 seats for the platform team this month.REQ-1041SSlack9,000 USDPurchase approval, heldMR
We need 50 GitHub seats. Please route whatever approval is required while the finance lead is away.REQ-1042GGitHub12,000 USDPurchase approval, heldHC
Need 30 Miro seats today; our team budget is already fully committed, so flag the owner.REQ-1043MMiro5,400 USDReassign, workingLB
We need eight Zoom seats, but I noticed the SOC 2 report has expired.REQ-1044ZZoom1,600 USDSecurity review, heldEW

Attached quote from Slack: please buy 20 seats for the platform team this month.REQ-1041Held

ApproveModifySend back More actions

State: Held
Assigned to: BOBudget owner (Finance)
Waiting: 7 h 38 min
Rule: R-SPEND-01
Supplier: Slack
Amount: 9,000 USD
Requester: MRMaya Rodriguez

SummaryWork record 11Evidence 5Decisions 1

Steps⌃

Record requestAgentCompleted

Reassign available seatsAgentReady

Purchase approvalBudget ownerHeldHeld under R-SPEND-01: 9,000 USD exceeds the 5,000 USD cap

Summary

PropertyNowProposed

SeatsNot provided20

VendorNot providedSlack

Amount0 USD9,000 USD

Contract term12 months12 months

Mandate

The agent acts. The code decides what it may do. The record proves it.

Independent AI product · in progress

AGENT SPEND

Partner OperationsDR-4471

Deal registrationsAssignment group is Partner Operations or Partner Escalations1 to 7 of 34

NumberShort descriptionStateAssignment groupUpdated

DR-4468Kestrel Medical Group, renewal expansionApprovedPartner Operations11:22
DR-4465Northlake Surgical, net-new registrationUnder ReviewPartner Escalations11:14
DR-4471Brightmoor Health, registered by Ardsley Technology GroupUnder ReviewPartner Operations11:05
DR-4462Vandermere Clinics, competitive overlapReturnedPartner Operations10:41
DR-4459Ashcroft Behavioral Health, platform expansionApprovedPartner Operations10:12
DR-4456Gilhaven Regional, tier upgrade requestUnder ReviewPartner Escalations09:38
DR-4453Merrow Point Physicians, new logoDraftPartner Operations09:04

DR-4471Deal registration for Brightmoor Health, registered by Ardsley Technology Group

ApproveSend back for detailEscalateMore actions

State: Under Review
Assignment group: Partner OperationsPartner Escalations
Assigned to: Priya RaghunathanTom Bergstrom
Account: Brightmoor Health
Channel partner: Ardsley Technology Group
Partner tier: Specialist

DetailsApprovals

Registration details

Prior returns on this account: Withheld from partnerHeld by escalation owner
Registered opportunity: OPP-11208
Margin band: Tier 2, standard
Review note: Overlaps an open registration

Related records5 of 5

TypeRecordDetails

Channel partnerUSER-88214Specialist tier
AccountACCT-3390Brightmoor Health
Registered opportunityOPP-11208Held by the internal reviewer
Implementation projectPRJ-2207Brightmoor Health System
Enablement recordENR-0413Two current certifications

Partner Operations

One partner request across five systems that had accumulated rather than composed.

Production product · recreated

PARTNER OPERATIONS

Three defaults the work didn’t survive.

Each was the reasonable default. The work showed what it cost.

A confidence thresholdExplicit review states
Lumens AI
A continuous score left the lawyer setting a threshold with no basis for choosing one. The concept replaced it with draft / reviewed, where review stays an explicit human act.
See the review states
One workflow for every siteTwo operating models
Patient Rounding
The mapping surfaced two different jobs. About a third of sites had a patient-experience team whose mandate was rounding; the rest had service-line leaders fitting it around everything else they owned. One needed depth, the other speed.
See the operating models
One status per requestLifecycle and custody move apart.
Partner Operations
A request that had changed hands still read as one status, so nobody could tell a stalled approval from a reassignment. The record now carries lifecycle and custody as separate tracks, and its own event log shows the state holding while responsibility moved twice.
See the custody ledger

Experiments

Smaller things, built to find something out.

Agentic research

Unpack

Can a research report stay honest after it ships?

IN MY DAILY WORKFLOW

Re-syncs against source data

Unpack.Retention brief · Q3Re-audit

1234

What changed after the rework

Setup-friction cancellations fell through August. Churn fell 34% (corrected from an initially claimed 38%) against Q24 and the cohort held.

The two accounts that churned anyway both cited pricing, not setup, in their exit calls.

Support-attributed churn

QuarterQ2Q3

Churn rate8.1%5.3%

Setup-flow exits319

Citation check · 4

Mar 4 customer sync · transcript

Figure differs from the sourceClaim re-synced to the transcript

Agent operations

Agent telemetry

Where does an always-on agent actually spend its money?

PROTOTYPE

Token spend tracking

Design tooling

Design context compiler

Does giving an agent the design system beat pasting a screenshot?

EVALUATION PROTOTYPE

Figma to design-token context

Working notes

Notes from the build.

AGENTIC UX

What "agentic UX" actually means

Underneath the buzzword sits a real design craft: shaping the boundary between what a system does on its own and what it checks with a person first.

Inside

Where does a feature sit on the autonomy line, and who decided?
Why "the AI can do it" is the beginning of the question, not the end
What makes a system act, ask, or hand control back
How do you make that boundary legible to the person on the other side?

What changes the answer

Reversibility
Cost of error
Confidence
Permissions
Who else is affected
Explain what it saw

The design work is deciding where a feature sits on this line, and making that legible to the person on the other side.

Rescheduling an internal focus block is not the same design problem as changing a patient priority or approving access to financial data. The interaction may look similar. The authority is not.

The design work isn't making the agent capable. It's making the boundary visible.

Covers autonomy, approval, reversibility, escalation and auditability.

PROCESS

The brief is the deliverable

Most of the design work that matters happens before any pixel exists. The decks come later. The brief is the deliverable.

On patient rounding, the critical decision was not the eventual screen. It was recognizing that the proposed product boundary did not match how rounding actually happened on the floor. Once that changed, the interface problem changed with it.

framing
product boundaries
assumptions
decision records

non-goals

Brief fragment

Problem boundary: A rounding feature scoped as a standalone build
User / role: Nurses rounding on the floor
Operational constraint: Logic already owned by the platform's existing sites
Product decision: Remap onto the existing site structure
Non-goal: A second surface beside the one nurses already use
Unresolved question: Still open

The earlier the ambiguity, the more expensive it is to carry forward.

A good brief doesn't remove the design work. It makes sure the team is solving the right problem.

PRODUCT

Enterprise AI is won in the workflow, not the model

Enterprise AI is won in the workflow around the model: who reviews what, where the system pauses, what it can access, and what happens when it gets one wrong.

The model is one dependency

Enterprise products also have permissions, policy, exceptions, queues, ownership, audit requirements, and existing systems of record. A better model does not make those disappear.

Model quality determines what is possible. Workflow design determines what is safe and useful enough to deploy.

Fig. 03 · Workflow anatomy

Swap the model and the product barely moves. Fix the workflow and it ships something people trust.

Aroundexceptionsrecovery

Onpermissionsapproval

Underprovenanceauditability

Fig. 04 · Three layers of control

The model determines what the system can do. The workflow determines when it should.

Compare notes

Email me