Architecture, internal systems, and applied AI
Essays and practical guides for anyone who feels the operation creaking: spreadsheets at their limit, systems that don't talk to each other, and processes that don't scale.
development costOld system: rewrite from scratch or replace it in parts?
A full rewrite puts all the risk into one cutover. Here is what the research says about replacing an old system in parts, and what coexistence costs.
9 min read
disconnected tool integrationWhen an integration retries, the order arrives twice
A timeout does not tell you whether the send worked. Here is why integrations duplicate orders, and the unique identifier that makes a retry safe.
7 min read
data security in applications (sixth pass: the account of someone who already left)Ex-employee accounts: who can still log in?
Offboarding happens in HR, but access lives across many systems. Without an account list tied to your people records, a departed employee's access stays on.
7 min read
team adoption of process change (fifth pass: the switch-off date for the old system)When to switch off the old system after go-live
While the old system stays on, it competes with the new one on every entry. What the research suggests and how to set a switch-off date for the old system.
7 min read
dashboards and management reporting (fourth pass: the metric that becomes a target and the distortion the dashboard cannot see)When a metric becomes a target, dashboards stop measuring
When a metric becomes a target, people optimize the number, not the operation. How the dashboard loses its meaning and how to design measurement to prevent it.
7 min read
spreadsheets to internal systems (fifth pass: the copy that circulates by email and the version nobody can name)Spreadsheets by email: which version is the right one?
Copies of a spreadsheet circulate by email and nobody knows which counts. Why the file became the system of record, and what changes with an owner and history.
7 min read
innovation initiative prioritization (fourth pass: the score that looks objective and the disagreement between scorers)Priority scores: why two people score differently
Two people give one initiative different scores because the criterion is ambiguous. Test your queue's disagreement in one afternoon before trusting it.
7 min read
process automation and ROI (fourth pass: the variants and volume that set the break-even point)How many variants does the process you will automate have?
Automation pays off depending on how often a process runs and how many variants it has. Here is how to count both before you decide what to automate.
8 min read
measuring innovation results and ROI (fourth pass: the comparison unit instead of before and after)Measuring innovation results: before and after is not enough
Before and after mixes the project's effect with what would have changed anyway. Use a comparison unit to prove the result of an innovation project.
8 min read
eliminating bottlenecks and repetitive tasks (fourth pass: the resumption cost the task timer does not measure)Interruption cost: what the task timer does not see
Timing only the execution of an ad hoc request understates its cost. In a 27-person study, resuming work took 11 to 17 minutes. The fix is at the entry point.
7 min read
how to scale the operation without bloating the team (native EN pauta: the designed fallback mode for when a system the operation depends on is down)When the system is down, what does your operation do?
An outage is a technology event. What the operation does next is an architecture decision. What CMC, AWS and hospital data show, and four decisions to make.
8 min read
development cost (fourth pass: the cost of changing a finished system and where the rule lives)Cost of changing software late: what the data shows
Does fixing late cost 100 times more? Across 171 projects the effect did not show up consistently. Our reading: the cost depends on where the rule lives.
7 min read
disconnected systems integration (fifth pass: the handoff between teams)Handoffs between teams: where the request gets lost
Requests get lost at the handoff between teams because context stays with the sender. Fix it with one record and explicit acceptance by the receiver.
7 min read
justifying technology investment to the board (fourth pass: the cost range instead of the single number)Tech project estimates: give the board a range
In a study of 1,471 IT projects, one in six overran cost by 200% on average. Take the board a cost range with review triggers, not a single number.
7 min read
disconnected tools integration (fourth pass: the age of the data a decision uses)How old can your data be when a decision uses it?
The right sync frequency between systems is the one that fits the decision the data feeds. Here is how to set a maximum acceptable age for each piece of data.
9 min read
scaling operations without bloating the team (fourth pass: the slack an always-full team does not have)An always-full team: the queue outgrows the load
Near 100% utilization, any peak becomes a queue and the wait grows faster than the load. Why a team's slack is designed into the operation, not hired.
7 min read
rapid prototyping (fourth pass: who tests the prototype, the operator and not only the approver)Who should test the prototype of an internal system?
The person who approves the budget is not the person who runs the system. Why five operators on real tasks find more problems than a demo for the board.
8 min read
hiring developers vs outsourcing to a studio (fourth pass: contract format and who carries the uncertainty)Fixed price or hourly: how to choose a software contract
Fixed price or hourly? It depends on how much of what you asked for is already decided. What the research shows and how to split the risk with your vendor.
9 min read
data security in applications (native EN pauta: vendor bank-detail changes as a system rule, not a habit)Who approves a vendor's new bank details?
A vendor's new bank details are a payment control, not an email task. What FBI, AFP and UK data show, and how to move the check from habit into the system.
7 min read
data security in applications (recovery that only exists once it has been tested)Does your backup restore? Only a restore test can say
Having a backup does not prove your company can recover. A restore test shows how much data you lose, how long you are down and in what order systems return.
7 min read
team adoption of process change (fourth pass: the line manager as owner of adoption)New system adoption: the line manager decides
Whether a team adopts a new system is decided by the line manager of the people who operate it. What Gallup and Prosci show, and how to design for it.
7 min read
dashboards and management reporting (third pass: recording the decision the panel prompted)Dashboards: record the decision, not just the number
Your management dashboard shows the number, but does it keep the decision that number prompted? Here is how to record what was decided and prove it works.
7 min read
spreadsheets to internal systems (fourth pass: the error the spreadsheet won't let you see)Spreadsheet errors: why nobody notices until they cost you
0.8% to 1.8% of formula cells in operational spreadsheets are wrong and nothing flags it. Why nobody sees it, and what changes when the rule leaves the cell.
7 min read
process automation and ROI (third pass: the system of record that makes return measurable)How to measure automation ROI without a system of record
Without a system of record, you have no 'before' cost to compare against. Here is how to measure automation ROI by starting with the data the process leaves.
8 min read
prioritizing innovation initiatives (third pass: the inventory that comes before the queue)Initiative inventory: you can only prioritize what you see
Before you prioritize, list everything already running: owner, cost and baseline for each initiative. Without an inventory, the queue reflects who asked.
8 min read
data security in applications (native EN pauta: the vendor breach notification clock across jurisdictions)When a vendor's breach becomes your legal deadline
A vendor breach still starts your own notification clock. Under GDPR, UK, Australian and US state rules, here is what needs to be ready before it rings.
9 min read
justify technology investment to the board (third pass: the cost of delay when the board says 'not now')Delaying a tech project has a cost. How much?
Putting a tech project on hold is not free. The problem bills you every month and the workarounds grow. Here is how to price the wait in your business case.
7 min read
hiring developers vs outsourcing to a studio (third pass: key-person dependency in the internal system)Does your internal system depend on one person?
When an internal system depends on one person, the risk is not that they leave. It is what exists only in their head. Four layers to move out before they go.
8 min read
measuring innovation results and ROI (third pass: the measurement window and the transition that contaminates the verdict)When to measure the result of an innovation project
Innovation projects get worse before they get better. When to open the measurement window, and what to record so the right project does not get killed.
8 min read
eliminating bottlenecks and repetitive tasks (third pass: human review of automated output as the new bottleneck)AI automation: the bottleneck is now the review queue
Automating a step does not remove the work. It returns as a review queue. How to size that queue and decide what actually needs a human check.
7 min read
cost of development (third pass: quote dispersion as information about scope)Software quotes: why three prices come back so far apart
Three quotes for the same internal system land far apart because each vendor read a different amount of coupling in your operation. What to ask first.
8 min read
scaling the operation without growing headcount (third pass: the team answering questions the system should answer)Self-service: the limit is the data, not the screen
Self-service does not fail at the interface. It fails when no system holds the state of the process. How to test that before you buy an answer layer.
8 min read
process automation and ROIAI automation cost: the bill comes per run
Rule-based automation charges once to build it. Put a model in the flow and the bill becomes per run, growing with the volume of your operation.
8 min read
team adoption of process change (third pass: the exception path the operation builds outside the system)Process exceptions: the path nobody logs
A process exception is a parallel path your operation builds and your system never records. How to measure that path and design the exception into the flow.
8 min read
build vs buy and the cost of leavingYour exit fee is disappearing. Your lock-in is not.
From 12 January 2027 a cloud provider cannot charge you to leave. The fee was never what kept you there. Four questions decide whether you can go.
9 min read
disconnected tool integrationYour automation stopped. When will you find out?
An automation that stops raises no alarm, it just stops running. Why silent failure is the default and the four decisions that give you the alarm back.
8 min read
rapid prototyping (third pass: sizing a pilot by case coverage, not duration)Innovation pilots: how many cases must yours see?
A pilot that worked may have seen only the common path. How to size a pilot by case coverage instead of calendar time before you approve the build.
9 min read
spreadsheets → internal systems (third pass: the operation that happens outside the office)Why field data only reaches you at the end of the day
The record of what happens in the field is created hours later, typed by someone who was not there. That is an architecture decision, not a signal problem.
8 min read
hiring developers vs contracting a studio (second pass: the in-house team that now writes code with AI)Building internal systems with AI: what changed
AI made writing code cheap, not the decisions that hold an internal system together. What that changes when you choose between building in-house and hiring.
7 min read
bubble → custom codeWhy AI agents get Bubble migrations wrong
An AI agent reading a Bubble export sees every page and workflow, not which ones matter. Why that breaks the rebuild, and what to settle before any code.
7 min read
prioritizing innovation initiatives (second pass: the concurrency limit per operational surface)How many simultaneous changes can your operation take?
The limit on simultaneous change is counted per operational surface, not per initiative. Here is how to map where your internal projects collide.
8 min read
systems architectureWhy AI stopped at marketing and not at operations
US and EU statistics agencies show AI landed where the output is language. Operations stayed manual because software cannot act on a state nobody wrote down.
8 min read
measuring innovation results and ROIWhy productivity gains never reach the bottom line
An hour saved is capacity, not cash. Here are the three conversions that turn a productivity gain into a business result, and what to decide first.
8 min read
scaling operations without growing headcountWhy your second location does not copy the first
Opening a second location copies the site and the staff, not the rule that made the first one work. What has to leave people's heads before you sign.
8 min read
bubble → custom codeBubble scaling limits: which ones actually bite
Bubble publishes its hard limits. Most never touch you. See which ones decide whether an operation can keep growing on the platform, and which are just noise.
6 min read
bubble → custom codeHow long a migration off Bubble actually takes
The schedule is set by parallel running, not by coding speed. See what decides the duration of a Bubble migration and why cutting over per workflow shortens it.
5 min read
bubble → custom codeMigrating from Bubble to custom code: how to decide
Leaving Bubble is a scoping problem, not a loyalty test. What the platform documents about ownership, what drives the rebuild cost, and how to run the move.
7 min read
bubble → custom codeWhat a Bubble rebuild costs, and what drives it
Nobody can price a Bubble rebuild from a screenshot. See the four things that set the number, and how to compare it against what staying already costs you.
6 min read
bubble → custom codeMigrating off Bubble: what transfers and what you rebuild
Your data leaves Bubble. Your logic does not. See what the platform documents about both, and how that split sets the real cost of migrating off it.
7 min read
rapid prototyping (second pass: promoting a validated prototype into production)Prototype approved: what's missing to make it a system
A prototype proved demand, not architecture. The four questions it never answered, and the criterion for promoting, rewriting or switching it off.
8 min read
process automation and ROI (third pass: AI autonomy level per flow)AI in operations: how much autonomy per process
The question is not which AI to use, it's how much autonomy to give it in each flow. The four levels, what each one costs, and when a fixed rule wins.
8 min read
prioritizing innovation initiatives (second pass: shutdown criteria and freeing managerial attention)When to shut down an innovation project that won't scale
Ending an initiative is an architecture decision, not a matter of courage. Define the stop criterion, the owner and the review date before approving it.
8 min read
cost of development (second pass: total cost after delivery)What it costs to keep a system running
A system's cost does not end at delivery. What you pay after go-live, and how the design decides that bill before the first line of code is written.
8 min read
disconnected tools (second pass: integrability as a buying criterion)How to evaluate integration before buying a tool
Before signing with a new software vendor, check the API, data export and cost of leaving. A six-point routine so you don't buy yourself an island.
6 min read
eliminating bottlenecks and repetitive work (second pass: the bottleneck is the wait between steps)Approval bottlenecks: where the process really stops
Slow approvals are the invisible bottleneck in an operation. How to measure time spent waiting in decision queues and design thresholds that free the flow.
7 min read
team buy-in for process change (second pass: measuring real adoption after go-live)System adoption: why go-live is not the end
Go-live is not adoption. Measure real usage of an internal system with four indicators, and know what to do when the team drifts back to spreadsheets.
8 min read
dashboards and management reporting (second pass: metric overload)Too many metrics: why measuring everything stalls you
A crowded dashboard blocks decisions instead of supporting them. How to cut metrics down, give every KPI an owner, and turn data into action again.
7 min read
spreadsheets to internal systems (second pass: how to migrate)Migrating off spreadsheets without stopping the company
A practical routine for moving from spreadsheets to an internal system without halting the operation: source of truth, flow-by-flow migration, dated cutover.
7 min read
disconnected tools (angle: shadow IT and stack governance)Shadow IT: when every department picks its own tool
Each department buys its own tool and the stack becomes a patchwork. Why banning doesn't work, and how to govern it without blocking innovation.
8 min read
data security in applicationsAccess control: who sees what in your systems
In 67% of security incidents the root cause is identity, not a technical flaw. How to structure access control across the systems your company runs.
6 min read
justifying technology investment to leadershipHow to justify a technology investment to your board
Boards don't reject technology, they reject risk nobody sized. How to build a case that starts from the cost of the problem and gets approved in phases.
6 min read
eliminating bottlenecks and repetitive workRepetitive tasks: what to eliminate first
Cutting repetitive tasks without looking at the flow wastes the effort. How to locate the real bottleneck, order the automation queue and measure the gain.
7 min read
disconnected systems integrationSystems integration: the cost of not integrating
Disconnected systems create rework, numbers that never reconcile and decisions made blind. What the lack of integration costs, and where to start fixing it.
6 min read
rapid prototypingRapid prototyping: validating before you invest
How to validate a system or automation idea in weeks, with a hypothesis, a success criterion and real data, before committing budget and people to it.
6 min read
dashboards and management reportingManagement dashboards: why the numbers don't match
Your dashboard disagrees with finance because each department defines the metric differently and nobody built a single source of data. How to fix the order.
7 min read
process automation and ROIDoes process automation pay off? How to calculate ROI
Automation pays when the flow is repetitive and was redesigned before it became code. How to calculate the payback and where the gain leaks away.
6 min read
prioritizing innovation initiativesHow to prioritize innovation and get past the pilot
Prioritizing innovation is not ranking ideas by impact and effort. It is deciding which bottleneck unblocks the others before you approve the pilot.
7 min read
team buy-in for process changeWhy your team isn't adopting the new internal system
Teams reject a system that was designed without the operation. Adoption is a result of architecture and involvement, not of training or pressure.
7 min read
measuring innovation results and ROIHow to measure innovation ROI and prove it upward
Innovation ROI usually stalls at board level for lack of method, not lack of return. How to set the baseline, attribute the gain and prove the value.
6 min read
spreadsheets → internal systemsWhen to replace spreadsheets with custom software
Spreadsheets stop scaling long before anyone admits it. See the five signs your operation has outgrown them, and how to decide on custom software without waste.
6 min read
automation in practiceHow to document and version corporate automations
What to record for each automation, how to version without sophisticated tooling, and how to make documenting a habit the team keeps after the first week.
7 min read
operations → controlStandardize critical processes without bureaucracy
How to standardize critical processes without freezing the operation: what to prioritize, how to document so people actually read it, and where automation pays.
7 min read
automation in practiceWhen manual processes and automation belong together
Practical criteria for deciding what to automate and what to keep manual: human judgment, exceptions, unstable rules, and what the data actually says about it.
7 min read
data → decisionData centralization: myths and truths as you grow
Myths and truths about centralizing data: what actually changes in decisions, what it costs, where to start, and the traps that make these projects expensive.
7 min read
operations → control7 ways to map hidden bottlenecks in your operation
Seven practical ways to find the bottlenecks no report shows: direct observation, process mapping, queue analysis, listening at the edges and small tests.
7 min read
data → decisionHow data quality prevents wrong business decisions
Duplicate, incomplete or stale data distorts reports and decisions. See the most common problems, what they cost, and how to keep data you can actually trust.
6 min read
automation in practiceOrchestrating IT and business teams in automation
How to align IT and business teams on automation projects: clear expectations, review rituals, the role of leadership, and what PMI and Deloitte data shows.
7 min read
operations → controlA monthly inspection checklist for internal systems
A monthly inspection checklist for internal systems: five blocks from backups to permissions and data, and how to keep the routine alive without bureaucracy.
7 min read
operations → controlAudit logs: how to analyze operational deviations
How to use audit logs to find operational deviations: the 5W1H structure, six analysis steps, and the route from a raw record to continuous improvement.
7 min read
systems architecture5 failures that break internal system integrations
The five failures that break internal system integrations: connecting before mapping, inconsistent data, weak governance, shallow tests and an unready team.
6 min read
automation in practice5 onboarding practices for enterprise applications
Five onboarding practices for enterprise applications: personalize by role, short steps, guided practice, support at the moment of doubt, and real feedback.
7 min read




