TokenOps

AI capital infrastructure

AI capital.Governed end to end.

Boards need one evidence base to decide where AI investment belongs, prove what it produces, and control how approved capabilities operate.

Explore the two problems See the Gymshark illustration
Board signal2026 director survey
44%

AI and technology risk or opportunity

42%

Technology adoption as a capital focus

38%

Enterprise AI deployment as a priority

01 — The board agenda

Two problems.
One operating reality.

AI is no longer only an innovation topic. It is simultaneously a capital-allocation decision and a production-control obligation.

1Investment

Capital control

Determine which AI investments to fund, how much capital to commit, and what evidence should justify continued spending.

ProductivityAutomationRevenue ProductTransformation
2Operations

Operational governance

Govern AI across accuracy, privacy, cybersecurity, intellectual property, compliance, data quality and autonomous action.

Effective controls. Credible economics. Clear accountability.

02 — The connection

One operational truth.

Capital control and runtime governance depend on the same precise understanding of the business.

01Where AI is used
02What work it performs
03What it costs
04What outcomes it produces
05What risks it creates

03 — Enterprise illustration

Gymshark.

A global, launch-intensive Shopify Plus business makes the connection visible: high-volume customer work, distributed data, defined policies, and actions that cross the line from assistance into production.

Customer-service and commerce work

Illustrative workflow surface
Product availability and sizingDelivery and returns Customer and order historiesPromotion and regional policy Inventory and fulfilment recordsApproved Shopify updates
01

Assisted work

People use models to accelerate tasks.

Usage is discovered across tools, teams and workflows.

02

Measured evidence

Economics and outcomes become visible.

Cost, adoption, impact and risk inform the investment decision.

03

Controlled production

Ready workflows move into governed runtime.

Rules, permissions, approvals and exceptions bound every action.

04 — Two views, shared telemetry

Decide with evidence.
Operate with control.

C

Capital view

TLACap

The allocation layer connects AI activity to cost, adoption and business value.

  1. Classify AI activity by investment type
  2. Attribute model, software, integration and labour costs
  3. Measure business impact and adoption
  4. Connect investment to productivity, revenue, margin, working capital or risk
  5. Identify consumption that is not producing sufficient value
R

Runtime view

milli.run

The operating layer implements only the workflow portions ready for controlled production.

  1. Govern access to Shopify and enterprise systems
  2. Enforce business rules and regional policies
  3. Restrict permitted actions and transaction values
  4. Require human approval at defined decision points
  5. Escalate exceptions and ambiguous cases
  6. Record every model decision, system action and human intervention
The combined model

Capital intelligence and runtime control form one learning system.

TLACap determines where AI capital creates value. milli.run ensures approved capabilities operate within defined controls. Production metrics flow back into the next investment decision.

01Allocate

Fund the right AI activity.

02Control

Bound production behavior.

03Measure

Return cost, risk and outcomes.

Return to the thesis

Point of view

Solving AI's Two Connected Problems

Download PDF

Capital control and operational governance require the same evidence base: a precise understanding of where AI is used, what it does, what it costs, what risks it creates, and what outcomes it produces.

AI has moved from an innovation topic to a board-level capital-allocation and governance issue. In a 2026 director survey, 44% of respondents prioritized AI and technology risks or opportunities for upcoming board meetings, 42% identified technology adoption as a major capital focus, and 38% ranked enterprise AI deployment among their top strategic priorities.

The Two Problems Boards Must Solve

Boards must solve capital control and operational governance together: deciding where investment belongs while ensuring approved AI capabilities operate inside clear production controls.

Why the Problems Are Connected

Capital control and operational governance depend on the same operational understanding of the business.

Boards cannot make sound investment decisions without knowing:

  • where AI is being used
  • what work it performs
  • what it costs
  • what outcomes it produces
  • what risks it creates

Similarly, a company cannot automate work responsibly without understanding the underlying workflow, the data involved, the actions the system may take, and where human judgment remains necessary.

The same operational data supports both responsibilities: capital control determines where investment should be made, while runtime governance determines how approved capabilities operate in production.

Illustration: Gymshark

Gymshark is a globally recognized fitness-apparel brand operating on Shopify Plus. Its scale, international reach, product-launch intensity, and high-volume customer interactions make it a useful illustration of how AI capital control and runtime governance connect in a larger enterprise setting.

Customer-Service and Commerce Work

Customer-service and commerce operations may include:

  • answering product availability, sizing, delivery, and return questions
  • summarizing customer, order, and interaction histories
  • drafting responses across email, chat, and social channels
  • interpreting promotion, returns, warranty, and regional policies
  • checking inventory, fulfillment, fraud, and sourcing records
  • preparing approved updates in Shopify and connected systems

Employees may initially use large language models to assist with these tasks, sometimes transferring data from Shopify, email, customer-service platforms, or internal records into a model. These activities do not represent a single AI investment. Some are productivity tools, some may develop into process automation, and others may support revenue growth, customer retention, or customer-facing capabilities.

The Capital View: TLACap

TLACap provides the capital-allocation view. It uses usage, identity, workflow, cost, and outcome data to:

  1. classify each AI activity by investment type
  2. attribute model, software, integration, and labor costs
  3. measure business impact and adoption
  4. determine whether the activity improves productivity, revenue, margin, working capital, or risk
  5. identify AI consumption that is not producing sufficient value

The same workflow evidence helps determine which activities are ready for production deployment. Routine order-status enquiries, approved address changes, standard return-eligibility checks, and policy-grounded product guidance may become sufficiently repeatable for controlled automation. Questions involving suspected fraud, disputed charges, exceptional refunds, safety issues, influencer or athlete relationships, or material reputational risk may remain human-assisted or require explicit approval.

The Runtime View: milli.run

A runtime such as milli.run implements only the portions of the workflow that are ready for production. It can:

  1. govern access to Shopify and connected enterprise systems
  2. enforce business rules and regional policies
  3. restrict permitted actions and transaction values
  4. require human approval at defined decision points
  5. escalate exceptions and ambiguous cases
  6. record each model decision, system action, and human intervention

The runtime then produces operational metrics, including workflow volume, completion rates, escalations, exceptions, latency, model usage, customer outcomes, and financial impact. TLACap uses those same metrics to assess cost, productivity, margin, capital deployment, and whether additional investment is justified.

The Combined Model

TLACap determines where AI capital is creating value. The runtime ensures that approved AI capabilities operate within defined controls. Both rely on the same underlying view of workflows, usage, cost, risk, and outcomes.

Source note: Shopify identifies Gymshark as a Shopify Plus merchant in its published customer case study, “How Gymshark outgrew Adobe Commerce and replatformed with Shopify Plus.”