An AI-powered competitive intelligence agent built in Cowork/Claude Code that monitors, analyses, and delivers actionable intelligence on Tide's competitive landscape — across both business banking/platform and acquiring.
Team: George Schmidt / Samson Muhangi Goal: Working prototype with real data.
| Competitor | Markets | Notes |
|---|---|---|
| Revolut Business | UK, EU | Broadest feature set, aggressive expansion |
| Starling Business | UK | Strong UK challenger, marketplace model |
| Monzo Business | UK | Consumer-first, growing B2B |
| HSBC Kinetic | UK | Incumbent digital play |
| NatWest Mettle | UK | Free business banking, incumbent-backed |
| Barclays Business | UK | Largest incumbent, digital push |
| Qonto | EU (FR, DE, ES, IT) | Leading EU SMB platform — key competitor for Tide's EU launch |
| RazorpayX | India | Full-stack business banking + payments, $9.2B valuation, IPO planned |
| Open (BankOpen) | India | SME neobank — current accounts, invoicing, payouts, expense management |
| NiYO | India | Business banking + employee benefits, strong blue-collar/SMB wedge |
| Competitor | Markets | Notes |
|---|---|---|
| SumUp | UK, EU | Low-cost card readers, expanding into business banking |
| Dojo | UK | Payments-first, strong in hospitality |
| Square | UK, EU | Ecosystem play (POS, banking, payroll) |
| Zettle (PayPal) | UK, EU | PayPal-backed, integrated with PayPal ecosystem |
| Worldpay | UK, EU, Global | Enterprise-scale, recently re-independent |
| myPOS | UK, EU | Hardware + acquiring, SMB focus |
| Tyl (NatWest) | UK | Incumbent acquiring play |
Total: 17 competitors across business banking/platform and acquiring (UK, EU, India)
- Web search for latest competitor news, product launches, funding, pricing changes
- Scrape competitor websites for feature and pricing data
- Process uploaded source material (PDFs, articles, press releases)
- Track job postings as leading indicators of strategic direction
- Structured competitor profiles — one file per competitor, consistently formatted
- Side-by-side feature and pricing comparisons (Tide vs. X)
- Gap analysis: what competitors have that Tide doesn't, and vice versa
- Positioning analysis: how competitors talk about themselves vs. how Tide does
- Weekly competitive digest — executive briefing format
- Alert on significant changes (new feature launch, pricing change, funding round, market entry)
- Product implications: "Competitor X launched Y — here's how it affects our roadmap"
- Marketing implications: "Competitor X is repositioning around Z — here's what that means for our messaging"
- All outputs are structured markdown — directly ingestible into Obsidian/knowledge base
[[wikilinks]]between competitor profiles, comparisons, and insights- Comparisons are versioned — run the same comparison monthly and track changes
- Exportable as PDF for stakeholder distribution
All outputs are structured markdown files designed to be:
- Ingestible — drop into any Obsidian/Zettelkasten knowledge base
- Linkable —
[[wikilinks]]connect competitors, comparisons, insights, and tasks - Versionable — re-run comparisons monthly, diff the changes over time
- Exportable — render as PDF for sharing with non-technical stakeholders
- Actionable — every output ends with implications and recommended actions
- Competitor Profiles (
competitors/) — structured company profiles with threat scores - Comparisons (
analysis/) — side-by-side feature/pricing/positioning comparisons - Weekly Digests (
reports/) — executive briefing of the week's competitive moves - Alerts — triggered when significant competitor changes are detected
- Tasks — action items extracted from analysis (e.g. "Investigate Qonto's new invoicing feature for parity")
competitor-intel/
├── CLAUDE.md ← Agent brain — instructions, role, rules
├── competitors/ ← One file per competitor (structured profiles)
│ ├── revolut-business.md
│ ├── starling-business.md
│ ├── monzo-business.md
│ ├── hsbc-kinetic.md
│ ├── natwest-mettle.md
│ ├── barclays-business.md
│ ├── qonto.md
│ ├── sumup.md
│ ├── dojo.md
│ ├── square.md
│ ├── zettle.md
│ ├── worldpay.md
│ ├── mypos.md
│ └── tyl.md
├── analysis/ ← Comparisons and gap analyses
│ ├── tide-vs-qonto.md
│ ├── tide-vs-starling.md
│ ├── acquiring-landscape.md
│ └── ...
├── reports/ ← Weekly digests and summaries
│ └── weekly-digest-2026-06-27.md
├── sources/ ← Uploaded PDFs, articles, raw material
├── tasks/ ← Action items from analysis
└── skills/ ← Reusable agent skills
├── scan-competitor.md ← Scan a single competitor for latest intel
├── weekly-digest.md ← Generate weekly competitive digest
├── compare.md ← Side-by-side comparison
└── threat-score.md ← Calculate/update threat scores
Goal: Project structure, agent brain, seeded competitor profiles
- Create project structure in workspace
- Write
CLAUDE.md— full agent instructions - Create competitor profile template
- Pre-research and seed all 17 competitor profiles with real data
- Define threat scoring model (weighted criteria)
- Calculate initial threat scores
Output: A populated knowledge base with 17 real competitor profiles
Goal: Working skills that pull real data and generate analysis
- Build
scan-competitorskill — web search + profile update - Build
compareskill — side-by-side comparison generator - Build
weekly-digestskill — executive briefing from all profiles - Build
threat-scoreskill — weighted scoring model - Test each skill with real data (Qonto, Starling as primary test cases)
- Verify outputs are well-formatted and knowledge-base-ready
Output: Four working skills that produce real, actionable intelligence
Goal: Demo-ready prototype with compelling outputs
- Generate Tide vs Qonto comparison (EU demo)
- Generate Tide vs Starling comparison (UK demo)
- Generate acquiring landscape comparison
- Create first weekly digest with real data
- Build threat score leaderboard
- Upload and process 2-3 real source documents (PDFs)
- Prepare 3-minute demo script
- Prepare ROI slide
- Dry run the demo — identify the live demo moment
Output: Complete demo-ready package
"Our product and marketing teams manually track ~14 competitors across business banking and acquiring. That intelligence is scattered across bookmarks, Slack threads, and people's heads. When a competitor launches something, we often find out days later — if at all. We're reactive when we should be proactive."
- Show the pre-built competitor profiles — real data, structured, linked
- Live moment: Run
scan-competitoron Qonto — agent searches the web, pulls latest news, updates the profile, flags implications for Tide's EU launch. Real-time, real data. - Show a pre-generated Tide vs Starling comparison — feature matrix, pricing, implications
- Show the threat score leaderboard
- Time saved: ~4-8 hrs/week manual research × multiple teams = 200-400 hrs/year
- Speed: Competitor moves surfaced in hours, not weeks
- Quality: Structured, comparable, compounding — not scattered notes
- Scale: This is a reusable skill. Any team can deploy it for their competitive landscape. Plugs into Google Workspace Intelligence + Gemini at org level.
- Next step: Integrate into product and marketing team workflows. Weekly automated digest. Monthly deep-dive comparisons.
| Metric | Without Agent | With Agent |
|---|---|---|
| Time to surface competitor move | Days to weeks | Hours |
| Research hours per week (per team) | 4-8 hours | ~30 min review |
| Annual time saved (per team) | — | 200-400 hours |
| Knowledge persistence | Scattered, ephemeral | Structured, compounding |
| Cross-team visibility | Siloed | Shared knowledge base |
| Intelligence quality | Inconsistent, reactive | Structured, proactive |
- Threat Leaderboard — competitors ranked by threat score (0-100), with movement arrows (↑↓→)
- Red Alerts — triggered when a competitor makes a move that directly threatens a Tide product line
- Weekly "One to Watch" — the digest highlights the single competitor that moved the most that week
- Workshop angle: "Whose agent surfaces the most actionable insight during the build session?"
- Environment: Cowork / Claude Code (terminal)
- Data privacy: Processing via Anthropic Enterprise — zero data retention, no training on inputs
- GWS connectors: Currently can't connect to Drive/Docs. Workaround: export docs as PDFs, upload to Cowork manually
- Sources: Web search (live), uploaded PDFs, manual captures
- Output format: Structured markdown → ingestible into Obsidian or any knowledge base → exportable as PDF
- Post-workshop: Integrate Claude Routines for scheduled weekly scans and digests
-
CLAUDE.md— full agent instructions - Competitor profile template
- All 14 competitor profiles pre-seeded with real data
- Skill definitions (scan, compare, digest, threat-score)
- Tide vs Qonto comparison (pre-generated)
- Tide vs Starling comparison (pre-generated)
- Threat score leaderboard
- Demo script (polished)
- ROI slide content
Plan created: 27 June 2026 Status: Ready for review before implementation