Creative problem-solving in business systems: when standard solutions don’t fit
You bought the “best practice” software, yet your team still keeps workarounds in spreadsheets and Slack. You’re not broken—your business is unique. The trick is knowing when to bend the system and when to design around it.
In this piece, I’ll show a simple way to apply creative problem-solving to business systems, with examples from real operations. We’ll cover how to reframe the problem, explore options beyond the obvious, and land on elegant fixes—often small, sometimes AI-powered—that snap into your existing tools (including SAP) without creating chaos.
Why “standard” breaks in the real world
- Standard processes assume a generic business. You deal in customers, suppliers, and edge cases with names and nuances.
- Integrated suites solve 80% well; the last 20%—pricing quirks, approvals, compliance edges—creates drag, rework, and frustrated teams.
- Mismatched processes quietly compound costs: context switching, duplicate data entry, manual reconciliations, slow handoffs, and hard-to-trace errors.
- The urgency now: AI, low-code, and APIs make bespoke fixes feasible for small teams—if you pick the right problem and design with intent.
The cost of forcing a “best practice” fit is rarely the license fee—it’s the everyday friction your people learn to live with.
A simple framework you can run in a week
Creative problem-solving (CPS) blends two modes: diverge to explore options, converge to decide what’s practical. Use this cadence:
- Frame the right problem
- Ask: What outcome is blocked? What’s the “job to be done” behind the workaround?
- Map the flow from trigger to value. Circle delays, handoffs, and rework.
- Challenge assumptions with “What if we eliminated this step entirely?”
- Diverge wide
- Generate options without judgment for 30–45 minutes.
- Use SCAMPER prompts: Substitute, Combine, Adapt, Modify, Put to another use, Eliminate, Reverse.
- Try role-switching: “If a customer designed this, how would it work?”
- Converge with constraints
- Score by effort, impact, risk, and time-to-value.
- Favor solutions that touch the fewest systems and change the fewest habits.
- Look for a “week-1 prototype” you can test with real users.
- Prototype, test, and iterate
- Ship a sketch: a form, a bot, a script, a one-field change in ERP.
- Capture feedback in 3 buckets: what to keep, what to adjust, what to drop.
- Decide to scale, tweak, or sunset. Document the decision.
Tip: A good candidate for automation typically occurs >30 times a week, takes >3 minutes, and has clear rules or patterns.
Five elegant workaround patterns that actually work
- Sidecar instead of teardown
- Pattern: Keep your core system stable. Handle exceptions in a small “sidecar” service or app that writes back via API.
- Example: Special pricing approvals sit in a lightweight web form that applies rules and posts the final price back to SAP/ERP. Sales gets speed; finance keeps control.
- Automate the “in-between”
- Pattern: The pain is often in moving context between systems, not in the systems themselves.
- Example: A short script or RPA flow reads orders every 10 minutes, classifies them by delivery risk with an AI model, and tags at-risk orders for human review in the existing dashboard.
- Human-in-the-loop AI
- Pattern: Let AI do first-draft work; people do the final mile.
- Examples:
- Customer emails: AI drafts responses from a policy library; agents approve or edit.
- Quality notes: AI summarizes defect photos and free text into structured codes for the MES/ERP.
- Why it works: Speed without blind automation; quality improves through feedback.
- Micro-app for the “last meter”
- Pattern: Build a tiny app for the frontline where friction lives. Keep it laser-focused on one job.
- Examples:
- A barcode-scanning progressive web app that posts receipts to SAP Business One and attaches a photo of the delivery note.
- A shop-floor “one-tap” downtime code selector that feeds OEE analytics.
- Decision support beats “big automation”
- Pattern: Many problems improve more from better decisions than from full automation.
- Examples:
- A daily exceptions list that flags orders missing key data, ranked by impact.
- A margin “guardrail” that pops an alert when a quote dips below target, with suggested fixes.
Real-world inspiration (and why it matters)
- Netflix reframed the delivery model (DVDs → streaming) and kept iterating on the user journey (e.g., trailers that auto-play) to reduce friction. Lesson: Rethink the path, not just the step.
- Uber removed two blockers—cash handling and trust—by going cashless and adding ratings. Lesson: Solve the human constraint, not just the logistics.
- GE Healthcare turned scary pediatric MRIs into “pirate adventures.” Lesson: Emotions are part of the system; design for them.
- IBM used design thinking to simplify a cloud platform and speed deployment. Lesson: Clarity and coherence beat feature bloat.
For small businesses, similar principles apply:
- Reframe the job: Is the goal “enter data” or “make a confident decision fast”?
- Remove blockers: If approvals stall, change the policy window and track exceptions.
- Design for emotion: Reduce cognitive load; make the next step obvious.
Three scenarios across different business types
-
Time-strapped professional (law, accounting, consulting)
- Problem: Intake and document prep steal billable time.
- Creative fix: An intake form that generates a case summary and a document checklist via AI, routed to the right matter in your DMS and calendar. Humans do the final check. Result: Fewer back-and-forths, faster starts.
-
Operations-focused owner (manufacturing, logistics, construction)
- Problem: Shop-floor data is late or messy; planning suffers.
- Creative fix: A scanner-first micro-app that captures scrap reasons and ties them to work orders. A daily exception list flags lines trending off target before MRP explodes.
-
Growth-minded entrepreneur (retail, hospitality, services)
- Problem: Promotions boost sales but crush margins or ops.
- Creative fix: A pricing “guardrail” that simulates margin impact by channel and inventory before the promo goes live, with prebuilt “safe” templates for staff.
Where AI and low-code fit (without the hype)
- AI shines at classification, summarization, and suggestion. Use it to draft, not decide, in regulated or high-stakes steps.
- Low-code/no-code is ideal for the “last meter” app: forms, approvals, simple data writes. Keep logic simple; shift complex rules to a service or your ERP.
- Integration matters: Prefer official APIs, OData services, or standard connectors for SAP and mainstream platforms. Avoid brittle screen-scraping unless it’s a short-term bridge.
Guardrails:
- Data security: Limit scopes, use service accounts, and log everything.
- Change control: Pilot with a single team; document and version your flows.
- Maintainability: One owner, one page of docs, and a rollback plan.
Quick-start CPS canvas
| Step | Questions to ask | Output |
|---|---|---|
| Problem | What outcome is blocked? Who feels it? What’s the measurable pain (time, errors, rework)? | One-sentence problem statement |
| Assumptions | What must be true today? What if those constraints were flexible? | List of challengeable assumptions |
| Ideas (diverge) | How might we eliminate, combine, or move steps earlier? What would a “draft-first” version look like? | 10–20 unfiltered ideas |
| Decide (converge) | What gives value in <2 weeks? What touches the fewest systems? What risk is acceptable? | Shortlist + pick one prototype |
| Pilot | How will we measure impact? What feedback loop and owner? | Mini spec + success criteria |
| Scale | What needs to be hardened (auth, logging, docs)? Who owns it? | Rollout plan + responsibilities |
A 10-day implementation plan
- Day 1: Frame the problem and success metric.
- Day 2: Map the current flow; collect 3 real examples.
- Day 3: Diverge—20 ideas in 45 minutes. No judgment.
- Day 4: Converge—score options; pick one prototype.
- Day 5–6: Build the smallest viable version (form/bot/script).
- Day 7: Test with 3–5 users on live but low-risk data.
- Day 8: Capture feedback; adjust once.
- Day 9: Decide to scale, tweak, or sunset. Document.
- Day 10: If scaling, add logging, simple training, and a rollback.
Common objections (and practical answers)
- “We’ll end up with a mess of mini-tools.”
- Set a sandbox, name things consistently, keep a catalog, and appoint an owner.
- “It won’t be supported by the vendor.”
- Stay within supported APIs/extension points. Keep the core clean; extend at the edges.
- “AI makes mistakes.”
- Keep humans in the loop. Log prompts/outputs. Use AI for draft/sort, not final decisions.
- “Change will slow us down.”
- Pilot in one team. Timebox to two weeks. Celebrate quick wins; deprecate what doesn’t work.
When not to customize
- Compliance or audit-critical flows that require strict vendor support.
- Areas the vendor roadmap will cover soon (confirm with your partner).
- Processes that are actually policy problems in disguise; fix the rule, not the software.
Key takeaways
- Your uniqueness is an asset. When standard software doesn’t fit, design around the edges—don’t rip out the core.
- Small, elegant fixes—sidecars, in-between automations, and human-in-the-loop AI—often deliver outsized gains.
- A structured CPS loop (frame, diverge, converge, pilot, scale) keeps creativity practical and risk controlled.
One clear next step
Pick one high-friction step that happens often and takes more than 3 minutes. Run the 10-day plan to prototype a fix. If you want a second set of eyes, outline the problem, constraints, and systems involved—I’m happy to suggest three right-sized options and a safe pilot path.
What becomes possible
When you stop forcing your business into a generic mold and start solving the right problems creatively, you get speed without chaos, control without bureaucracy, and systems that finally work the way your people do. That’s not just efficiency—it’s a competitive edge.