Search

Independent publication · First essays in preparation

Considered positions on enterprise AI, from the architecture up.

Out of True is an independent publication of technical analysis on AI in the enterprise: the models and agents, the data and infrastructure beneath them, and the economics that decide what they are worth. Written by Jay Meshram, a practising technical architect.

What Out of True covers About Jay
An independent publication on enterprise AIReading −3.4° · scroll to bring it true

Why “Out of True”

In engineering, a structure is out of true when it no longer sits square to its reference line. It may still stand. The work is measuring the gap and deciding what it means. In enterprise AI, four gaps are worth measuring.

Promise and evidenceWhat a capability is said to do, set against what it has been shown to do.
Intent and behaviourHow an architecture was meant to work, set against how the running system behaves.
Investment and valueWhere the budget goes, set against where the return turns up.
Assumption and possibilityWhat the established patterns take for granted, set against what has recently become feasible.

Sometimes the gap is a warning. Often it is an opening.

What it covers

The whole system, not just the model

Most consequential questions in enterprise AI sit where one layer meets another: where a model meets real data, where an agent meets an existing process, where a pilot meets a budget. That is where Out of True looks.

The perspective

Architecture as the lens, not the limit

Whatever the subject, a model release, an agent framework or a funding decision, the question is the same. What does it change in how systems are designed, run and paid for, and what does it leave untouched?

A request passes through an agent to a model, data and tools. The intended path runs straight to the intended outcome; the observed path bends through the tools and lands below it.RequestAgentModelDataToolsIntendedObserved
IntendedObserved

Desk 1

Models and capabilities

What new model capabilities actually change for enterprise systems, and which architectural assumptions they retire or leave standing.

Desk 2

Agents and applications

Agents, copilots and AI inside existing applications: where autonomy earns its place, and what it costs to operate.

Desk 3

Data, integration and infrastructure

Retrieval, grounding, permissions, integration and the platforms underneath: the layers that decide whether an answer can be trusted.

Desk 4

Economics and operating models

Cost, value, evaluation and the organisational shape AI programmes need, from pilot economics to who owns the system once it runs.

The question is rarely whether AI works. It is where, how, and at what cost.

The rest is detail, and the detail is the work

What to expect

Every piece works through four questions.

Some pieces argue. Some take a system apart. Some revisit an earlier position. All of them answer these.

  1. What is changing?

    The development itself, described precisely enough to separate what is new from what has been renamed.

  2. Which assumptions does it touch?

    The technical and architectural beliefs that held until now, and which of them no longer do.

  3. What are the trade-offs?

    The costs, the failure modes and the reasonable readings that point the other way.

  4. What should enterprises reconsider?

    A position on what follows: what to change, what to wait on, and what to leave alone.

Who writes it

Jay Meshram, technical architect

I began in instrumentation and control, commissioning DCS and SCADA systems, where out of true is a measurement rather than a metaphor. A transmitter reads a fraction off, the board stays green, and the plant does something nobody designed it to do.

I have spent the decade since designing enterprise platforms on Salesforce, and now work as a technical architect on Agentforce and Data 360. Out of True is where I write independently about what I see across the wider field.

More about Jay

At a glance

Based
Toronto
Practice
AI and Enterprise Architecture
Started in
Instrumentation and control systems
Writing
Independently, in a personal capacity