Start With Friction, Not Technology

The RFP asked which LLM we preferred.

Nobody asked which task wasted forty hours a week.

Technology first.

Friction never.

That meeting order predicts project outcome.

The Opinion

Most automation projects begin with technology.

The best automation projects begin with friction.

Friction is the drag between knowing what to do and doing it.

Searching.

Copying.

Reconciling.

Escalating.

Re-explaining.

Waiting.

Repeating.

Technology without friction diagnosis builds solutions searching for problems.

Problems without friction language get funded as AI pilots instead of workflow fixes.

Marketplace and ecommerce teams feel this pattern constantly.

New vendor demo every quarter.

Same manual spreadsheet every Monday.

Starting with friction reverses the sequence.

Name the drag.

Quantify the drag.

Then choose tool, automation, or process fix.

See The Operational Friction Score™.

Wrong Questions

These questions sound strategic.

They often skip diagnosis.

Should we use OpenAI?

Vendor question.

Not workflow question.

Should we use Claude?

Model comparison without task definition.

Should we build an agent?

Architecture fantasy before path mapping.

Should we buy a platform?

Platform search replaces friction inventory.

Wrong questions produce funded experiments.

Experiments without friction maps rarely reach production.

See Why Most AI Projects Never Reach Production.

Better Questions

Better questions sound mundane.

That is why they work.

What repetitive work exists?

List tasks by frequency times duration.

What decisions repeat?

Identify decision types with same criteria weekly.

What workflows create friction?

Score information, workflow, decision, system, ownership friction.

See The Cost of Operational Friction.

Who waits on whom?

Coordination tax sources.

See The Coordination Tax.

What breaks when volume doubles?

Scaling exposes friction debt.

Better questions produce investment targets.

Targets may need no new technology.

Friction Before AI

AI magnifies whatever exists.

Clear friction map with weak model sometimes beats fuzzy map with strong model.

Example:

Operators copy suppression data from email to spreadsheet to case tool.

Friction is duplicate entry and context loss.

Wrong fix: AI summary of emails.

Better fix: suppression queue with structured fields feeding case creation.

Maybe AI assists case draft inside queue.

Maybe not yet.

Queue removes more friction than model choice.

See AI Is Not a System.

Marketplace Friction Examples

Suppression workflow

Friction: status checked in three places.

Fix: single queue with detection feed.

Technology choice secondary.

Weekly reporting

Friction: six manual exports and paste.

Fix: scheduled pull to source of truth.

Not new dashboard color scheme.

Case management

Friction: assignment via Slack ping.

Fix: owner required on queue entry with revenue rank.

Not AI case writer first.

Catalog updates

Friction: attribute fixes in sheet then manual upload.

Fix: write-back integration from queue.

Automation follows path clarity.

See The Hidden Cost of Spreadsheet-Based Operations.

Spreadsheets often signal friction accumulation, not solution.

Decision Friction vs Tool Shopping

Teams debate tools while decision friction remains.

Standup re-debates priority daily.

Leadership buys prioritization software without sort rules.

Software displays random order beautifully.

Friction remains.

Fix decision friction with layers.

See The Revenue-at-Risk Framework™.

Then software encodes sort.

Sequence matters.

Friction Inventory Exercise

One week exercise for any team.

Each operator logs friction events:

What slowed you?

How long?

How often per week?

Category: information, workflow, decision, system, ownership.

Sum hours.

Multiply by loaded cost.

That number is business case.

Compare to tool subscription cost.

Often friction removal via process beats tool purchase.

Sometimes tool purchase wins after friction mapped precisely.

Mapping prevents vague transformation budget.

See The Operational Debt Framework™.

Unpaid friction becomes debt with interest.

When Technology Is the Answer

Technology first is correct sometimes.

After friction inventory proves:

Integration gap is primary drag.

Manual step cannot be removed by process simplification.

Volume exceeds human review capacity with clear criteria.

Data exists but access path is broken.

Even then, specify friction removed.

Not technology adopted.

Example: automated suppression detection removes detection friction measured in hours per day.

Not we adopted AI platform.

Language shapes project success.

AI After Friction

When friction inventory points to language or classification repetition:

Embed AI at that step.

Inside workflow.

With review gate.

With write-back.

See The Difference Between a Prompt and a Process.

See The Journey From Prompt to Process to Software (And Why Most Teams Stop Too Early).

Friction-first sequencing prevents prompt theater.

Leadership Habit Change

Replace opening question in automation meetings.

Old: what tool should we use?

New: what friction hours do we remove this quarter?

Old: are competitors using AI?

New: which workflow still runs on copy-paste?

Old: can we demo something this month?

New: can we measure drag reduction in ninety days?

Boring questions.

Better outcomes.

See The Best Software I’ve Built Was Boring.

Closing Thought

Start with friction.

Not technology.

Name repetitive work.

Name repeated decisions.

Name drag categories.

Quantify hours.

Fix highest drag with simplest intervention.

Sometimes intervention is integration.

Sometimes process.

Sometimes ownership rule.

Sometimes AI at one step.

Technology choice comes last.

Outcomes come first.

Organizations that invert the sequence keep buying tools while operators keep copying spreadsheets.

Organizations that start with friction build boring systems that compound.

Choose friction first.

That is operator discipline vendors will never sell you.

Reference this opinion at every automation kickoff.

No friction map, no project charter.

That rule alone would kill half of orphaned pilots.

The other half would finally reach production.

Both outcomes improve.

Start there.

Friction Maps in Cross-Functional Work

Cross-functional friction is often coordination tax wearing workflow clothes.

Merchandising and marketplace ops both touch catalog rows.

Both run separate status meetings because no shared queue exists.

Friction map reveals meeting as tax source.

Fix is shared row-level visibility with named owners.

Not better meeting agenda.

See Every Business Runs on Four Systems.

Information and decision systems weak create coordination meetings as workaround.

Fix systems.

Meetings shrink.

The RFP Problem

RFPs that start with technology vendor comparison skip friction inventory.

Vendors answer beautifully.

Operators still copy spreadsheets.

RFP should request friction hours removed per quarter and integration scope for write-back.

Model comparison belongs after friction map and workflow diagram exist.

See The Difference Between a Prompt and a Process.

RFP language shapes vendor responses.

Ask for workflow embedding.

Vendors propose embedding.

Ask for model only.

Vendors propose tabs.

Words matter before contracts signed.

Peak Season and Friction

Peak season multiplies friction hours.

Manual steps that cost ten minutes daily cost twenty under fatigue.

Approval chains that worked at normal volume create backlog spikes.

Friction map before peak identifies which steps to remove before volume arrives.

Removing steps in peak is harder than removing steps in calm month.

See The Cost of Waiting: Why Operational Delays Compound.

Friction removed in calm month pays interest during peak.

Friction ignored in calm month collects debt during peak.

Plan accordingly.

Many friction items resolve without AI.

Remove duplicate spreadsheet.

Require owner on entry.

Apply revenue-at-risk sort.

Automate one export.

Integrate two systems.

Retire one approval step.

Each fix boring.

Each fix measurable.

AI can wait until language repetition remains after boring fixes applied.

See The Best Software I’ve Built Was Boring.

Boring fixes fund trust for later AI steps.

Skipping boring fixes and jumping to AI produces demos on broken paths.

Final Word

Technology is not the starting point.

Friction is.

Describe the problem clearly before automating it.

Quantify drag before selecting vendors.

Fix highest drag with simplest intervention.

Embed AI only where repetition remains after path is clean.

That sequence is not anti-innovation.

It is pro-outcomes.

Innovation without friction diagnosis is expensive guessing.

Friction diagnosis with modest tools is operational engineering.

Choose engineering.

Reference this opinion at every kickoff.

No friction map, no charter.

That single rule changes project portfolios faster than any model upgrade announcement.

Start with friction.

Everything else follows.

Document the Friction Map

Friction map should be one page per workflow.

Workflow name.

Volume per week.

Hours per event by friction category.

Named owner for map update.

Review date monthly.

Attach map to project charter.

Charter without map is vendor shopping.

Map without charter is complaint session.

Together they become engineering brief.

Engineering brief produces boring software that reaches production.

See The Best Software I’ve Built Was Boring.

Boring software starts with friction map honesty.

Flash software starts with vendor demo memory.

Choose starting point deliberately.

Deliberate starts beat enthusiastic starts for operations teams tired of abandoned pilots.

Respect that fatigue.

Friction map is respect translated into method.

Use it.

Agents Are Not Exempt

Agent language does not bypass friction law.

Autonomous agent still needs trigger, input schema, review gate, write-back, owner, and metric.

Agent without those is autonomous demo.

Marketplace volume destroys autonomous demo quickly.

See Most AI Projects Fail Before the AI Matters.

Friction map first.

Agent second.

Same sequence as any other tool label.

Labels change.

Sequence does not.

Hold sequence constant across hype cycles.

Operators trust constants.

Trust funds production.

Production funds compound outcomes.

Start with friction.

Always.

Regardless of headline.

Headlines rotate.

Friction math persists.

Build on math.

Closing Summary

Friction-first automation means describing drag before selecting tools.

It means quantifying hours before approving budget.

It means fixing duplicate steps before debating model vendor.

It means embedding AI inside clean workflows instead of opening tabs on broken paths.

Technology choice is the last question.

Friction is the first.

Organizations that invert the question order keep funding experiments that operators abandon under volume.

Organizations that hold the order build systems operators defend during peak season.

Defended systems compound.

Abandoned experiments decay.

Choose compound path.

Start with friction.

Document it.

Fix highest drag.

Measure reduction.

Then ask which tool, if any, remains necessary.

Often the answer surprises leadership.

Often the answer is integration, not intelligence.

That surprise is the opinion doing its job.

Reference this article when the first meeting question is which AI vendor to pick.

Reply with which friction hours the vendor removes with write-back proof.

Meeting tone changes immediately.

That change is worth the article alone.

Start with friction.

Not technology.

Always.