# GCC Atlas — Product thesis Prototype: https://gcc-atlas.arkapravo.in ## 1. GCC Atlas A decision workspace for India GCC setup Speaker notes & sources: Product proposal. Desk research and a working prototype; no customer interviews or commercial validation have been completed. --- ## 2. The problem is a decision without comparable evidence A new entrant must commit to a location and partners before it has a reliable way to compare them. ### The decision Where can we build the capability we need, at a cost we understand, with partners we can verify? ### The fragmented inputs Company announcements, office reports, policy documents, recruitment estimates and referrals describe different slices of the ecosystem. ### The resulting friction The buyer rebuilds the evidence in a spreadsheet, resolves conflicting definitions, and asks advisors to fill the gaps. This is a synthesis of the assignment and desk research. The frequency and cost of this friction still need buyer interviews. Speaker notes & sources: Problem supplied in assignment. Inference, not measured customer behaviour. EY GCC Pulse Survey 2025 highlights talent and operating priorities: https://www.ey.com/en_in/insights/consulting/global-capability-centers/india-s-gccs-are-leading-the-shift-to-intelligent-ai-native-enterprises --- ## 3. The ecosystem has several information bottlenecks | Player | What they need | What they use today / likely consequence | | --- | --- | --- | | Entrant CTO / COO | Comparable city and setup evidence | Advisors, referrals, reports → repeated diligence | | GCC / HR leaders | Role-level hiring and capability data | Recruiters, peers → uneven salary / supply estimates | | Advisors, lawyers, auditors | Reliable entity and project context | Filings, client records → manual reconciliation | | Recruiters, EOR / payroll firms | Demand and operating requirements | Networks, job posts → weak project qualification | | Real estate firms | Mandates, space demand and timing | Leasing data, relationships → partial demand visibility | | States / investment agencies | Project eligibility and policy uptake | Applications, promotion → selective information | Stakeholder pain map = hypotheses to validate. Exposure to a fee or commercial incentive does not establish bias in any specific provider. Speaker notes & sources: Stakeholders drawn from assignment. Current workarounds and consequences are hypotheses based on desk synthesis, not primary interviews. --- ## 4. The numbers need a unit and a reporting period ### 1,700+ — GCCs · FY24 Nasscom–Zinnov 2,975+ centres ### 2,117 — GCCs · FY26 Nasscom–Zinnov 3,728 units ### 1,950+ — “GCC units” · Dec 2024 JLL office report Scope not reconciled A later year can explain growth. A different definition can explain a different count. Atlas preserves both; it does not average them. Speaker notes & sources: FY24 indexed Nasscom page: https://community.nasscom.in/communities/nasscom-insights/india-gcc-landscape-report-5-year-journey (direct fetch returned 406; indexed summary used). FY26 primary Zinnov release: https://www.prnewswire.com/in/news-releases/india-s-gccs-are-increasingly-leading-the-ai-mandate-for-global-enterprises-driving-global-value-creation-nasscom--zinnov-report-302817463.html . JLL: https://www.jll.com/en-in/newsroom/gccs-drive-record-77-2-mn-sqft-office-leasing-in-india . Do not treat their units as methodologically identical. --- ## 5. Small errors can matter more than cheaper office space Illustrative 200-person GCC: ₹24 lakh annual salary per person, 70% seating, 100 sq ft per seat, ₹110 base rent / sq ft / month. ### ₹48 cr — Annual salary budget 200 × ₹24 lakh Excludes benefits and hiring ### ₹2.4 cr — Effect of a 5% salary error ₹48 cr × 5% Sensitivity, not observed loss ### ₹18.5 L — Saving from 10% lower rent ₹1.848 cr base rent × 10% Same office assumptions The product should make assumptions explicit and demand role-level talent evidence. A rent-only ranking would answer the wrong question. Speaker notes & sources: All inputs are illustrative author assumptions. Salary = 200 * 24 * 100000 = 480000000 INR. Area = 200 * .7 * 100 = 14000 sq ft. Rent = 14000 * 110 * 12 = 18480000 INR. 5% salary = 24000000 INR. 10% rent = 1848000 INR. Not a market salary or rent benchmark. --- ## 6. The first opportunity is a defensible location shortlist | Opportunity | Value / data feasibility | Priority and rationale | | --- | --- | --- | | Location decision workspace | High decision stakes; some public precedents | 1 · First product, with explicit evidence gaps | | Launch / expansion intelligence | Recurring value; public announcements | 2 · Adds freshness and repeat use | | Provider delivery evidence | High trust value; public proof is thin | 3 · Build through consented references | | Talent / effective cost benchmarks | Critical input; data needs partners | 4 · License / collect; do not invent | | Full ecosystem marketplace | Broad demand; cold start and conflicts | Later · Requires verified supply and demand | Priority is an editorial judgment based on urgency, public-data feasibility and ability to test a specific buyer outcome. Speaker notes & sources: Prioritization is author judgment, not a numerical market survey. Provider delivery and talent information often require interviews, licenses or client consent beyond public documents. --- ## 7. The first buyer has an approved India mandate Initial segment: a global mid-market company planning its first 100–500-person technology GCC, before signing a lease or appointing a setup partner. ### User & buyer COO or transformation lead runs the comparison. CTO validates the capability. CFO signs the business case. ### Trigger Leadership approves India exploration. The company has a mandate and a staffing plan but little local context. ### Job to be done Produce a three-city shortlist with evidence, cost assumptions and unresolved questions that leadership can review. Entry point: the shortlist decision. Success means the buyer can explain why a city advances and what must still be verified. Speaker notes & sources: ICP, roles and trigger are explicit hypotheses. Sample scenario uses 200 staff; headcount range is a proposed target segment, not an established market distribution. --- ## 8. GCC Atlas: a working evidence workspace 6 cities 10 parent companies 16 parent–city records 21 linked sources Compare a scenario, inspect evidence and export a brief. A curated research sample, not a national census. Public / publisher claims are not an independent audit. Speaker notes & sources: Screenshot of authored local working prototype at http://127.0.0.1:4173. Counts from prototype/src/data.js. Seven parents meet the prototype precedent eligibility rule; three are retained with insufficient or old evidence. --- ## 9. The workflow produces a reviewable decision brief ### 01 Define the project Choose a mandate and sector. Enter headcount, salary and seating assumptions. Adjust peer versus cost weighting. ### 02 Challenge the shortlist Compare six cities. Open each company’s source. Keep announcement-only or uncertain mandate evidence separate. ### 03 Carry the reasoning forward Save up to three cities. Add diligence questions. Export Markdown or project JSON; print a PDF. Working today: ranking, keyword retrieval, database saves, optional accounts and export. Live crawling and runtime LLM generation remain outside the prototype. Speaker notes & sources: Feature claims reflect implemented app. Ranking uses matching eligible parent count within the sample and modeled annual salary + base rent. No city-specific talent benchmark is claimed. --- ## 10. Trust lives at the claim level A primary source can establish one fact without establishing every fact on the same company profile. ### Entity ≠ location One parent, multiple city records. Customer-facing offices are not automatically included as GCCs. ### Announcement ≠ operation Vanguard’s 2024 plan is retained as announcement-only. The sample does not infer a later opening from the planned date. ### Presence ≠ capability supply Fidelity’s office listing confirms a location. It does not establish the captive mandate there or available talent. Store source URL, publication period, access date, claim, evidence status and limitations. “Unknown” is a useful answer. Speaker notes & sources: Vanguard source: https://externalcrewnet.vanguard.com/content/corporatesite/us/en/corp/who-we-are/pressroom/press-release-vanguard-announces-a-new-technology-office-in-hyderabad-india-121924.html . Fidelity source: https://careers.fidelityinternational.com/locations/india/offices/ . Prototype does not claim Vanguard is still unlaunched; it says the selected evidence is insufficient to verify subsequent opening. --- ## 11. Existing players already offer data and execution | Alternative | What it already does | Atlas’s proposed reason to switch / add | | --- | --- | --- | | Nasscom / Zinnov reports | Market landscape and periodic research | Project-specific comparison with preserved scope | | GCCPROs | Database, city atlas, verified-field claims | Shared decision history and unresolved claims | | Scutiger tracker | Sourced launch / expansion tracking | Explicit assumptions and a buyer-owned brief | | ANSR / Zinnov services | GCC setup, operations and transformation | Independent comparison before partner selection | | Reports + spreadsheet | Flexible analysis with familiar tools | Evidence attached to each claim; less reassembly | Positioning: an evidence workspace for GCC setup decisions. Sourced rows alone are not a moat; workflow adoption and correction history must earn one. Speaker notes & sources: GCCPROs: https://www.gccpros.com/about . Scutiger: https://www.scutiger.com/gcc-tracker/ . ANSR: https://ansr.com/global-capability-center/ . Zinnov: https://zinnov.com/gcc-setup/ . Research report links in slide 4. No claim that these competitors cannot implement the proposed workflow. --- ## 12. Each stakeholder needs a different reason to return | Stakeholder | Reason to use Atlas | Contribution / commercial boundary | | --- | --- | --- | | Entrant / GCC leader | One brief with assumptions and sources | Pays for decision workspace or diligence | | Advisor / legal / audit | Reusable evidence with a review trail | Contributes corrections; can license workspace | | Recruiter / payroll / EOR | Clear mandate and qualified requirements | Opt-in project access; no contact resale | | Real estate firm | Named space requirements and timing | Shares current quotes with date and scope | | Service provider | Documented relevant case evidence | Submits references; cannot buy rank | | State / investment agency | Correct policy links and eligibility questions | Publishes official updates; no outcome scoring | The entrant comes first. Other stakeholders join when they make the buyer’s evidence more useful, not merely to increase listings. Speaker notes & sources: Stakeholder acquisition and contributions are a product proposal. No partnerships, integrations, user interviews or commercial agreements have been secured. --- ## 13. A paid decision project fits episodic demand ### First distribution Recruit 5 entrant teams through GCC communities and independent advisors. Offer a scoped shortlist review using their actual mandate. ### Pricing tests Test ₹50,000 per project for the workspace. Test ₹2 lakh for concierge diligence. Prices and willingness to pay are unvalidated. ### Repeat use After setup, expansion alerts and a reusable advisor workspace can support subscriptions. Keep referral commissions out of ranking. Illustrative service scenario: 50 concierge projects × ₹2 lakh = ₹1 crore gross revenue, before delivery costs. This is not a TAM estimate or forecast. Speaker notes & sources: All pricing, conversion, volume and revenue are explicit hypotheses. The total GCC ecosystem revenue is not treated as addressable software revenue. Concierge gross-margin feasibility requires measuring analyst hours, licensing and acquisition costs. --- ## 14. AI accelerates extraction; review controls the claims Proposed production pipeline. The one-day prototype uses AI-assisted research and a curated static snapshot. ### 01 Collect & extract Monitor permitted public sources. Capture date and document version. Use a schema-guided model to extract claims with source spans. ### 02 Resolve & review Link parent, subsidiary and location. Flag conflicts and duplicate announcements. A reviewer approves identity and operating status. ### 03 Retrieve & explain Retrieve approved evidence for a project. Calculate costs in code. Generate a brief with citations and abstain when support is missing. Start with 50 double-reviewed records. Evaluate identity matches, source support and announcement/operation errors before expanding coverage. Speaker notes & sources: Production design proposal; not implemented integrations. Prototype uses deterministic keyword search and calculations, no runtime LLM. Source licensing and site terms must be assessed before ongoing ingestion. Claims retain effective period and observed date; revisions form correction history. --- ## 15. The next 30 days should test value before scale | Test | Success threshold (proposed) | Decision if it fails | | --- | --- | --- | | 10 discovery interviews | Recurring shortlist pain in ≥6; exact workaround recorded | Narrow the buyer or problem | | 5 project walkthroughs | ≥4 finish a sourced shortlist in 15 minutes | Simplify workflow / missing evidence | | 3 paid concierge pilots | ≥2 pay; contribution margin positive | Revisit pricing and service scope | | 50-record blind audit | ≥95% fields supported; zero false operating-status promotions | Fix evidence rules before ingestion | | Provider reference trial | Consent-based scope / outcomes for 5 cases | Defer delivery ratings | North-star hypothesis: approved project shortlists whose key claims are supported and whose open questions are explicit. No interviews or pilots have run yet. Speaker notes & sources: Thresholds are author-proposed decision gates, not reported results. Interview outreach is not part of this completed assignment and no messages were sent. Customer validation has not occurred. --- ## 16. One day of execution, with a deliberate scope ### Built Responsive evidence workspace, scenario comparison, company filters, source ledger, partner check plan and exportable brief. ### Research boundary Company and publisher evidence with original links. Small sample; no audited census, salary dataset, verified provider outcomes or automatic refresh. ### AI disclosure AI supported research, idea critique, initial schema, code and deck creation. Deterministic retrieval and scoring run in the app; no runtime LLM. Human judgment retained: the first buyer, exclusions, source interpretation, commercial neutrality and what the prototype must not claim. Speaker notes & sources: AI used extensively during this session to research and build. No fabricated user interviews, performance reviews or private source data. Browser review and tests are recorded separately in outputs/Validation-Report.md. --- ## 17. Try one location decision 1 Open City shortlist and choose Engineering. 2 Switch to Operations / BFSI; inspect the changed peers. 3 Inspect Fidelity or Vanguard in Company evidence. 4 Search “GCC units” in Evidence room. 5 Save three cities and download the decision brief. The core product promise: the reviewer can inspect the evidence behind the decision and see what is still unknown. Speaker notes & sources: Prototype: https://gcc-atlas.arkapravo.in . Deck: https://gcc-atlas.arkapravo.in/deck.html . PDF: https://gcc-atlas.arkapravo.in/GCC-Atlas-Deck.pdf . All cost inputs remain illustrative. Demo projects stay in the visiting browser; optional accounts continue across devices. --- ## 18. Research references Primary company, government and publisher sources. Full source ledger is available in the prototype; citations also appear in speaker notes. Research date: 11 October 2026. Sources establish their stated claims; they do not constitute independent verification of delivery or current staffing. Speaker notes & sources: All source URLs in prototype/src/data.js and outputs/Research-Ledger.md. Company evidence includes Sonatype, Vanguard, Sanofi, Mastercard, BNY, NatWest, JPMorganChase, Tesco, Walmart and Fidelity International. Publication dates may be unknown; access dates are not freshness guarantees.