Articles
Each one is dated, and the manufacturing pieces come first. Holden’s opinion pieces are in the full list, labeled as commentary.
Manufacturing
These five cover plant systems that disagree about what a part is, testing AI against a physical process, choosing the first project, and the controls an agent needs before it touches real work.
How we work with manufacturersIndustrial AI succeeds or fails on whether equipment, sensor, control, operator and business systems can hand each other a fact they agree on. A smarter model does nothing for a plant where they can’t, so the fix is a vocabulary the plant owns.
No outside body is testing AI against your production line, and NIST’s AITE testbed isn’t pointed at manufacturing. So you build the acceptance test: a frozen baseline, a read-only run across a full cycle, and a number written down before you look.
Score each candidate on business value, workflow fit, available evidence, measurable outcomes, risk and ownership before you pick one, and leave novelty out of it.
An AI opportunity assessment should produce measured workflows, ranked opportunities, a no-build list, and an actionable first-project brief.
Eight controls to check before an AI agent runs real business work, from a bounded job and an accountable owner to safe retries, monitoring and lifecycle management.
Any organization
Read these for what an agent is, which actions need a person’s approval, what a system costs to keep running, why agents drift without a maintained standard, and how to map a workflow first.
What agentic AI is, how it differs from chatbots and workflows, what its parts do, where it fails, and how to judge a real business use case.
Approval gets set workflow by workflow. It stays permanent on anything that can’t be recovered and loosens everywhere else as the agent earns trust, the way an employee does. Here’s the loop I’d run.
A real agent runs by itself with little downtime and still shows a net gain after maintenance, review, confusion, mistakes, and recovery are subtracted. Here is the bill most deployments never run, and the boring week-long test I would apply before anybody calls it ROI.
Agents drift when all they have is auto memory and one context file, and a single file has a hard ceiling. The fix is a maintained Operating Standard that already knows your organization, and it comes with a maintenance bill of its own.
Map the trigger, inputs, steps, exceptions, owner, output, and current cost of one workflow before deciding what an AI agent should do.
For non-profits, start with Own the Donor Record, Not the Dashboard and Adoption Is Not Capability: What Your Nonprofit's AI Number Actually Measures.
Archive
Opinion pieces are marked Commentary, and website pieces are marked Websites.
Seven months after opening Claude Code for the first time, my business runs on plain text files, 34 scheduled jobs and four rules. Here’s the machine, the rules file it runs on, and how to start yours in about an hour.
No outside body is testing AI against your production line, and NIST’s AITE testbed isn’t pointed at manufacturing. So you build the acceptance test: a frozen baseline, a read-only run across a full cycle, and a number written down before you look.
Measure AI agent ROI after review, exceptions, maintenance, failures, recovery, provider cost, and the value of useful new capability.
Nearly every nonprofit reports using AI and almost none report mission impact, because the sector is measuring private habits and calling them capability. Here’s the difference, and the four things that close it.
Industrial AI succeeds or fails on whether equipment, sensor, control, operator and business systems can hand each other a fact they agree on. A smarter model does nothing for a plant where they can’t, so the fix is a vocabulary the plant owns.
Map the trigger, inputs, steps, exceptions, owner, output, and current cost of one workflow before deciding what an AI agent should do.
Agents drift when all they have is auto memory and one context file, and a single file has a hard ceiling. The fix is a maintained Operating Standard that already knows your organization, and it comes with a maintenance bill of its own.
Agentic fundraising showed up as a category in one week. The question worth asking is who holds the donor record while you use it, because if the context lives only in the platform’s agent, switching platforms becomes amnesia.
Approval gets set workflow by workflow. It stays permanent on anything that can’t be recovered and loosens everywhere else as the agent earns trust, the way an employee does. Here’s the loop I’d run.
Build the Operating Standard alongside one useful workflow, so real work defines the context, controls, evidence, and reusable foundation.
A real agent runs by itself with little downtime and still shows a net gain after maintenance, review, confusion, mistakes, and recovery are subtracted. Here is the bill most deployments never run, and the boring week-long test I would apply before anybody calls it ROI.
An AI opportunity assessment should produce measured workflows, ranked opportunities, a no-build list, and an actionable first-project brief.
Eight controls to check before an AI agent runs real business work, from a bounded job and an accountable owner to safe retries, monitoring and lifecycle management.
A modern business website should explain the business deeply, capture real inquiries, preserve search equity, and stay current after launch, kept that way by the people who run the business.
MIT found 95% of organizations piloting generative AI saw zero measurable P&L return. Here’s what the research from MIT, McKinsey, RAND, Gartner, S&P Global and BCG says about how corporate AI programs die, and what the successful 5% do differently.
What agentic AI is, how it differs from chatbots and workflows, what its parts do, where it fails, and how to judge a real business use case.
Score each candidate on business value, workflow fit, available evidence, measurable outcomes, risk and ownership before you pick one, and leave novelty out of it.
If one of these articles sounds like your week, tell us which process it’s in, and we’ll tell you whether it’s worth working on.