Make the FTSE 100 Stack Play Nicely

Look around the FTSE and you see the same pattern: most organisations now run on a common spine — Microsoft (M365/Azure), SAP, Oracle, AWS, Google Cloud, Salesforce, Adobe, ServiceNow, Workday, Snowflake. Around the edge: Atlassian, Okta, Splunk. In FS/telco, Pega shows up a lot for complex casework and real-time decisioning.

This isn’t a hunch. Vendors publicly report strong FTSE penetration (e.g., ServiceNow “more than half”, Workday “40%+”, Oracle “majority”), and UK market studies show Azure and AWS out in front for cloud, with GCP third. Elsewhere the pattern shows in practice: SAP for ERP/finance, Microsoft almost everywhere for productivity, Salesforce/Adobe in the front office, Snowflake powering the data plane. Different histories; the same backbone.

*Illustrative, not exhaustive.

Most stacks don’t fail in flames — they just slow everything down. You feel it in the basics: two “sources of truth” for customer, dual writes that never reconcile, brittle links hard-wired to vendor hubs, identity and observability you don’t actually own. Then come the “quick tweaks” that block upgrades, DR that lives on slides not in code, and “differentiation” buried inside suites where it can’t earn. Result: lock-in, slower change, rising cost, flat impact.

What actually moves the needle?

• Decide who owns what. Write down, in plain English, what each platform owns, what it supports, and what it will never do. Name the system of record for customer, product, employee, asset. One SoR per domain. No dual writes. (Typical split: ServiceNow for ITSM/ESM & employee workflows; Pega for complex case/BPM and, where chosen, decisioning; Salesforce/Adobe for channel journeys; Snowflake as the analytical source of reference.)
• Make the seams simple — and yours. APIs for sync, events for distribution, CDC/batch for analytics. Keep identity central. Version everything. DR is code, not slides. Don’t hard-wire to vendor interfaces: end every integration on your façade (OpenAPI/AsyncAPI) and keep vendor specifics in adapters you control. Hold keys, logs/metrics/traces, and the broker on your side. Give breaking changes 90 days’ notice. Allow 6–12 months’ parallel run for exit. Track a simple Switchability score (0–5); Tier-1 platforms should be ≥3/5.
• Configure before you customise. Use documented extension points. Protect upgrades with playbooks and SLOs. If a “quick tweak” breaks upgrades, it wasn’t quick.
• Differentiate with a thin, portable layer. Your edge is proprietary signals, decision services (rules/ML) exposed as APIs, and journey orchestration across channels. Keep it loose-coupled so it survives a platform swap. Measure by P&L impact, not features shipped.

Wrap-up

Some closing thoughts:

• The stack is real. Make it yours: assign owners, define platform roles, keep seams neutral and portable. Put differentiation only where it pays—and measure it.
• Accept the stack. Design the game: who owns what, what each platform does, and neutral, portable seams. Put your scarce genius where it earns.
• The stack is given. Control the edges—clear ownership, explicit roles, neutral/portable seams. Place differentiation where it returns value.
• This is the stack. Lead it: name owners, fix roles, keep seams neutral and portable. Invest in differentiation only when it pays back.
• Take the stack as fact. Architect the rest: ownership, roles, portable seams. Concentrate differentiation where it pays—track the dividends.
• Reality is the stack. Governance is the lever: owners set, roles clear, seams neutral/portable. Put differentiation where it earns its keep.