Start with one process

From friction to flow.

A practical field guide for SMEs improving operations with better processes, software and AI. Twenty-two pages on finding the work worth fixing, choosing the right level of technology and proving the change, without launching a transformation programme.

Read the guide Print the guide (PDF, 22 pages)

From Friction to Flow, a field guide for SME leaders. Free to read here and free to print. No form in front of either.

  • 22pages
  • 5moves, friction to flow
  • 1workflow to try them on
  • Sep
    2026    published
The cover of Prodro’s field guide, From Friction to Flow.

Chapter one · The case for operational leverage · 9 minutes

Small firms do not need digital theatre.

They need important work to become faster, clearer, more reliable and easier to scale. Where UK AI adoption stands, why technology projects disappoint, and the five principles that keep a project tied to the operating result.

The operational advantage is not having more software. It is making the right work easier to repeat.

Kieran McCloud, Director, Prodro

Small and medium-sized firms are full of capable people compensating for imperfect systems. They remember the exception, copy the information, chase the handover and fix the spreadsheet. The business keeps moving, but its best people become the integration layer.

Software and AI can change that. But buying a tool before understanding the work often relocates the friction instead of removing it. The team gets another login, the data fragments again, and the promised benefit becomes difficult to prove.

Our approach starts with the operating reality: what triggers the work, who needs the outcome, where it stalls, what judgement is involved and what a better result would be worth. Only then do we decide whether the answer is a clearer process, better use of an existing system, integration, automation, AI or purpose-built software.

This guide is our practical method for making that decision. It is written for firms without enterprise transformation teams, unlimited budgets or time to waste.

AI use is rising. Operational depth is not.

The opportunity for SMEs is real, but the route to value has less to do with access to tools than with choosing and embedding the right use case.

UK businesses reporting use of at least one AI technology, June 2026 ONS [1]

  1. 0–9 staff28%
  2. 10 or more staff35%
  3. 250 or more staff49%

Up from around 12% for firms with 10 or more staff in September 2023. The smallest firms adopt last and least.

1.6

AI technologies used on average by an adopting business, up from 1.4 in 2023. Adoption has widened far faster than it has deepened.

ONS [1]
39%

cited difficulty identifying activities or business use cases as a barrier to adoption.

ONS 2023 research [2]

The bottleneck is not awareness. It is turning a broad technology possibility into a specific operational improvement that people can adopt and leaders can measure.

The OECD also identifies cost, relevance and trust in vendors as important barriers for UK SMEs [3]. A credible partner therefore has to be willing to recommend what not to buy, not only what to implement.

When the tool becomes the strategy.

The failure is rarely dramatic. It is usually a slow accumulation of workarounds, exceptions and unclear ownership. Trust in the decision falls a step at a time.

Trust in the decision, stage by stage.
  1. 01

    A tool is selected

    The category looks promising, but the operating problem is still broad.

  2. 02

    The process bends around it

    People adapt locally while the end-to-end workflow remains unchanged.

  3. 03

    Exceptions multiply

    Manual checks and spreadsheets reappear at the boundaries.

  4. 04

    The benefit becomes vague

    Usage is measured, but time, quality, margin and customer impact are not.

  5. 05

    Trust falls

    The next technology decision becomes harder, because the last one never proved itself.

The order that avoids it

  1. Understand the work
  2. Simplify the process
  3. Define the outcome
  4. Fit the technology
  5. Test in operation
  6. Scale only what proves useful

Five principles for practical digital improvement.

These principles keep a project tied to the operating result rather than the novelty of the technology.

  1. 01

    Start with friction, not features

    Find the repeated delay, error, handover, workaround or missed opportunity before discussing a solution.

  2. 02

    Fix the rule before automating the task

    If nobody agrees what should happen, software can only make the disagreement faster and harder to see.

  3. 03

    Use the lowest sufficient level of technology

    A template, a configuration change or an integration may create more value than a custom build.

  4. 04

    Keep human judgement where consequence is high

    AI can prepare, classify and recommend. Accountability stays with a named person.

  5. 05

    Prove value in the live workflow

    A prototype is useful only when it tests behaviour, operational fit and a measurable outcome.

If the right answer is a £20 tool you can buy tomorrow, that’s the answer we’ll give you.

Chapter two · The Friction-to-Flow method · 14 minutes

Five moves from operational friction to measurable flow.

The method stays technology-neutral until the work, the value and the constraints are understood. It is the working detail inside the four stages Prodro takes every engagement through.

  1. 1

    Understand

    See how the operation actually works, not how the process says it works.

  2. 2

    Decide

    Determine what should change and why.

  3. 3

    Deliver

    Put the agreed changes into practice, with clear ownership.

  4. 4

    Prove

    Measure what improved: time, cost, capacity or quality.

  1. 01Find frictionSee where the work stalls, repeats or relies on memory.
  2. 02Quantify impactTranslate annoyance into time, cost, quality, risk or growth.
  3. 03Redesign the workRemove waste and clarify rules, ownership and exceptions.
  4. 04Fit the technologyChoose from the six decisions.
  5. 05Prove and scaleTest the new flow with users, metrics and a decision date.

Move through the method quickly on a small workflow. Depth should follow evidence, not precede it.

Find the real friction.

Observe the workflow as it happens. The visible complaint is usually downstream from the actual constraint.

StageUnderstand

Ask

Start and finish

Name the trigger, the final outcome and the people who touch the work.

Do

Watch the work

Use screen-sharing and recent cases rather than an idealised process map.

Working notes

Watch forJumping to the most vocal complaint, or assuming every manual step is waste.

Leave this move with

A specific workflow, an observable constraint, and the people affected by it.

Quantify the value.

Convert friction into a business case that can survive a budget conversation. The calculator on prodro.co.uk does this arithmetic for one workflow.

StageUnderstand

Ask

What changes?

Time, delay, rework, risk, conversion, capacity, customer experience or decision quality?

Do

Build a baseline

Use a small sample if necessary, but define the denominator and the current state.

Working notes

Watch forInvented ROI, or treating every saved minute as cash that immediately leaves the cost base.

Leave this move with

A baseline, a target range, and one primary outcome metric.

Redesign the work.

Remove unnecessary complexity before asking technology to carry it.

StageDecide

Ask

What is essential?

Separate the purpose of the workflow from the habits and exceptions accumulated around it.

Do

Make it explicit

Define the common path, the exception path, decision rights, data needed and ownership.

Working notes

Watch forDigitising every current step, preserving redundant approvals, or ignoring the people who absorb the exceptions.

Leave this move with

A simpler target workflow that people recognise and can follow.

Fit the technology.

Choose software and AI in proportion to the job, the risk and the strategic value.

StageDecide

Ask

What capability is missing?

Recording, coordinating, integrating, automating, interpreting, or enabling a distinctive service?

Do

Name the decision

Each finding is answered by one of six decisions. Work out which one this is, and why.

Working notes

Watch forChoosing a platform because it is fashionable, over-engineering a low-value workflow, or underestimating support.

Leave this move with

A justified decision, with clear boundaries, owners and dependencies.

Prove and scale.

Put the better flow into live work, and use evidence to decide what happens next.

StageDeliver + Prove

Ask

What would convince us?

Define acceptable quality, adoption and business outcome before the pilot begins.

Do

Run the smallest live test

Include the people who do the work, the exceptions they see, monitoring and a fixed decision date.

Working notes

Watch forDeclaring success because a demo works, or scaling before the team can operate and support the new flow.

Leave this move with

A scale, revise or stop decision, backed by operational evidence.

Chapter three · Software and AI in the real world · 9 minutes

Every finding is answered by one of six decisions.

These are not rungs to climb. Each has its own trigger, and the work is naming which one a finding is. Where AI earns its place and where it does not, and seven questions to answer before it touches a live workflow.

  1. 1

    Stop

    The work serves no useful purpose. Remove it.

  2. 2

    Simplify

    The process is more complicated than the job requires. Redesign it.

  3. 3

    Buy

    Someone has already solved this properly. Use their product.

  4. 4

    Automate

    Repetitive human work that technology can absorb.

  5. 5

    Connect

    The problem exists because systems don’t talk to each other.

  6. 6

    Build Last resort

    The right solution doesn’t exist. Create it.

Most operational problems cross people, process and technology. Fixing one layer often just moves the work somewhere else.

AI is most useful between messy input and structured action.

It can read, classify, extract, draft, summarise and recommend. The business still needs a process, an owner and a definition of acceptable output.

Signals that AI may help

  • High-volume text, documents, images or conversations
  • A clear next action after interpretation
  • Outputs that can be reviewed or reversed
  • Enough examples to test quality and edge cases
  • A measurable reduction in delay, effort or inconsistency

Signals to pause

  • A broken or disputed process
  • High-consequence decisions with no accountable reviewer
  • Sensitive data in unapproved tools
  • No baseline or accepted quality threshold
  • A vague goal to “use AI” rather than improve an outcome

The pattern that holds

  1. AI prepares
  2. A person reviews
  3. Software records or triggers the approved action
  4. The team monitors exceptions and quality

As confidence and evidence grow, the human checkpoint can move. It should never disappear merely because the demo looked convincing.

AI should reduce the cost of good judgement, not remove responsibility for the outcome.

Seven questions before AI touches a live workflow.

Use stronger controls as consequence, sensitivity and autonomy increase. Government guidance also emphasises ongoing risk management throughout the AI lifecycle [6][7].

For a small firm, governance should be visible and usable: an approved-tools list, clear data rules, a named owner, test cases, an incident route and a regular review.

Before AI touches a live workflow0
Your ticks stay in this browser.Print the list (page 16 of the PDF)

Chapter four · Put it into practice · 8 minutes

Three scenarios, ninety days, one page.

Three SME scenarios, a 90-day path from decision to evidence, and the Process Opportunity Canvas: one page that takes a single workflow from its name to a pilot decision.

The same three layers, three different businesses.

Each scenario resolves to a process change, a software change and a bounded use of AI. Compare them side by side rather than one after another.

The friction

Partners rewrite similar material, pricing arrives late, and nobody can see why proposals stall.

A 30-person advisory firm wants faster proposals without turning every bid into a custom project. The symptom is slow drafting. The constraint is fragmented knowledge, unclear approval rules and manual handovers.

Process

Standardise the offer

Define reusable service modules, pricing rules, evidence requirements and approval thresholds.

Software

Create one flow

Connect CRM data, proposal templates, pricing and approval status in the existing stack.

AI

Draft with context

Use approved source material to create a first draft; a named lead reviews claims, scope and price.

Measure the change
  • Lead-to-proposal time
  • Manual touches per proposal
  • Conversion and gross-margin quality

Do not start withA general chatbot with access to every client file, or a custom platform before the offer and approval rules are stable.

The three scenarios are hypothetical composites created to explain the method. They are not Prodro client case studies and contain no claimed savings.

A small improvement should produce a decision within 90 days.

The objective is not to finish transformation. It is to prove whether one improved workflow deserves wider investment.

  1. Days 1–15

    See the work

    Observe the workflow, interview the people in it, establish the baseline and identify the constraint.

  2. Days 16–30

    Design the better flow

    Remove steps, define rules and exceptions, assign ownership, and name the decision: stop, simplify, buy, automate, connect or build.

  3. Days 31–60

    Pilot in operation

    Implement the smallest live intervention, train users and keep a visible exception route.

  4. Days 61–90

    Measure and decide

    Compare the outcome with the baseline. Scale, revise, stop, or move to the next constraint.

This 90-day route covers implementation and pilot activity. Prodro’s Start with one process engagement is a separate two-week diagnostic. It ends with the current state, the better state and an action plan you can act on with or without us. Implementation is not included.

Take one workflow from its name to a pilot decision.

Complete this before discussing vendors, platforms or features. Seven boxes, the same seven as page 20 of the print edition.

01WorkflowName the start event and the completed outcome.
02OutcomeWhat needs to be true when the workflow finishes?
03FrictionWhere does it wait, repeat, fail, depend on memory or require re-entry?
04BaselineVolume, time, cost, quality, risk or conversion today.
05Root causeProcess, ownership, information, capability or technology?
06The decisionStop, Simplify, Buy, Automate, Connect or Build?
07 · Pilot decision

What will we test, who owns it, which metric should move, and on what date will we decide?

Fill it in on screen, print it for the room, or send yourself a copy.

The online canvas keeps everything you type in your browser until you choose to send it.

Open the canvasPrint the guide (PDF, 22 pages)

Sources

The guide uses current UK evidence to frame the adoption challenge. The method, principles, scenarios and recommendations are Prodro’s interpretation.

  1. Office for National Statistics, Artificial intelligence in UK businesses: 2023 to 2026, 20 July 2026.ons.gov.uk/businessindustryandtrade/business/businessservices/articles/artificialintelligenceinukbusinesses/2023to2026
  2. Office for National Statistics, Management practices and the adoption of technology and artificial intelligence in UK firms: 2023, 24 March 2025.ons.gov.uk/economy/economicoutputandproductivity/productivitymeasures/articles/managementpracticesandtheadoptionoftechnologyandartificialintelligenceinukfirms2023/2025-03-24
  3. OECD, SME Technology Adoption in the United Kingdom, 22 April 2026.oecd.org/en/publications/2026/04/sme-technology-adoption-in-the-united-kingdom_4cba1e43.html
  4. Department for Business and Trade, Understanding technology adoption among UK SMEs, 31 July 2025.gov.uk/government/publications/understanding-technology-adoption-among-uk-smes
  5. Department for Business and Trade, SME Digital Adoption Taskforce: final report, 31 July 2025.gov.uk/government/publications/sme-digital-adoption-taskforce-final-report
  6. Department for Science, Innovation and Technology, Responsible AI Toolkit, updated 15 November 2024.gov.uk/government/collections/responsible-ai-toolkit
  7. Department for Science, Innovation and Technology, AI Risk Management Toolkit: guidance, 8 September 2026.gov.uk/government/publications/ai-risk-management-toolkit/ai-risk-management-toolkit-guidance

Scenarios. The three scenarios are hypothetical composites created to explain the method. They are not represented as Prodro client case studies and contain no claimed savings.

Scope. Published September 2026. Review legal, regulatory, data-protection and sector-specific requirements for each implementation.

© 2026 Prodro Group Limited. Registered in England and Wales, company no. 17268651. Registered office: 71–75 Shelton Street, Covent Garden, London WC2H 9JQ.

The guide gets you to a decision on paper. We can do it inside your operation.

If you would rather not take one apart yourself, our two-week engagement does it inside your operation. One defined workflow, 14 elapsed days, £4,500 fixed. You come away with how the work happens today, how it should, and a plan you can act on with or without us.

Credited against any implementation booked with Prodro within three months, up to the value of that work.

See the two-week engagement