The tax nobody voted for
Each tool solved a problem. Together they created a costlier one that appears on no invoice: the cost of fragmentation.
ReadBlog
Operating on your own logic
Essays in seasons, meant to be read in order. The first goes from symptom to cause in the fragmented company. The second goes down to the infrastructure a company needs to put AI to work.
The scenes in these episodes are composites of common situations, not customer cases. Any figures in them are illustrative.
Season 1
One version of the truth
What a fragmented company costs, why neither SaaS, nor the ERP, nor AI solves it on their own, and why your business logic is the asset worth protecting.
- Excel isn't the villain. It's the witness.The spreadsheet that carried the company through its early years now exposes a risk: critical rules without controls and operations that depend on them.
- The company that depended on one personIf the process lives in someone's head, the company doesn't own its operations. That person does.
- Five versions of the truthMeetings end up arguing about numbers instead of decisions. Sometimes the data is wrong; sometimes nobody wrote down the definitions.
- The SaaS that won't let you growIt solved your year-one problem. By year five, it decides how your team works.
- Having an ERP isn't a transformationMany companies celebrated their ERP go-live. Months later, they still operate as if it didn't exist.
- The AI agent mirageThe demo impressed the board. Three months later, the pilot still hasn't touched operations. The model needs real-case testing, and the data, permissions and process need preparation.
- Data doesn't create value. Decisions do.Forty dashboards published, and leadership still asks for the report in Excel. Before buying another BI tool, review the data, definitions and decisions.
- SaaS or your own platform: the decision that shapes the next decadeFor twenty years the question was buy or build. There's a third answer, and it changes the math.
- The advantage that doesn't come with a licenseSeason finale. Competitors can buy the same tools. The advantage depends on how you combine them with your rules, data and ability to execute.
Season 2
A company ready for AI
The technical part: semantic layers, infrastructure for agents, MCP, autonomy, evaluation, security, traceability and cost. And the decision to build it in-house or with a partner.
AGI and your business: surf the wave or get swept under
Nobody knows when artificial general intelligence will arrive, or whether it will arrive as imagined. What a company can assess is how ready it is to test each new capability.
Read- One truth for people and machinesA semantic layer shares business definitions across reports and agents. It reduces calculation errors when sources, permissions and questions are governed too.
- What an agent needs to workAn agent doesn't work in a vacuum. Connecting it to operations means defining how it queries, how it acts and which controls it needs; the specific pieces depend on the task.
- MCP: superpowers inside your companyMCP standardizes how an assistant discovers and uses capabilities. It includes optional HTTP authorization; business permissions, approvals and logging still need their own design.
- Graduated autonomy: where the agent decides and where a person doesThe question isn't whether to trust an agent, but how much, for what and with what safety net. Autonomy is designed in levels, like self-driving cars.
- How to evaluate an agent before trusting it with operationsAn agent demo is not a production test. Evaluating its results, failures and limits is a craft with a method.
- Security for agents: identity, least privilege and prompt injectionAn agent with access to your systems is a new kind of user, one that reads everything it's given and sometimes obeys instructions it shouldn't. Protecting it starts with fifty-year-old principles.
- If it can't be audited, it doesn't go to productionEvery figure an agent delivers and every action it takes should be traceable to its origin. Traceability isn't bureaucracy: it's what makes trust possible.
- The real cost of AI in productionThe pilot cost little. In production, model calls, tools, retries and human review all count. Measure value per case resolved.
- Build in-house or with a partnerSeason finale. IT spends the day fighting fires and has no dedicated team to build this infrastructure. The decision isn't ideological: it's about capacity, time and ownership.
Do any of these episodes sound familiar?
In 20 minutes we review your case and you leave with a map of what stays, what connects and what's worth replacing.