AI's Evolutionary March in Finance: Growth, Costs, and Data Readiness Define the Race
AI is set to lift sales and profits for financial institutions by up to 10% over the next few years, according to new research from Bloomberg Intelligence. Early gains are revenue-led-client engagement, personalization, cross-sell-while cost savings lag.
The catch: higher operating costs in the near term. Firms are spending big on data infrastructure, cloud, tooling, and model governance. Build capability now, see stronger margin benefits after 2027-2028 as productivity compounds and spend normalizes.
What the data says
- About 70% of financial firms expect 1-10% revenue growth from AI in 2-3 years; a similar share expects profit lift.
- Over 70% also expect higher operating costs tied to AI programs.
- Only 23% report enterprise-wide scaling of AI (vs. 35% across all industries).
- Survey base: 600+ leaders; 150 in financial services across insurance, asset/wealth management, commercial and investment banking, and fintech/payments. Majority are large enterprises, with North America and Europe dominant.
Why costs rise before savings
- Data pipelines: cleanup, labeling, lineage, and entitlements
- Cloud programs: migration, security hardening, and observability
- Model lifecycle: evaluation, monitoring, and model risk management
- Controls: privacy tooling, PII handling, audit trails, explainability
In short: the early phase is about capability build, not immediate cost takeout.
Headcount is set to rise
Two-thirds of executives expect headcount growth over the next three years due to AI. This is workflow realignment, not a staff cull.
- New roles: AI compliance specialists, workflow and agent designers, model auditors
- Adjacencies: data product managers, prompt/interaction designers, AI SRE/ops
Why finance scales slower than retail and tech
- Tight regulation, high model-risk standards, and required explainability
- Fragmented legacy data and strict PII controls
- Higher cost of model development, validation, and continuous monitoring
Where budgets are flowing
Payments leads AI allocation, spending about 17% of total tech budgets. Investment banking and wealth management follow.
Commercial flows are the big prize. Mastercard estimates roughly $80T in commercial flows, with $63T in invoiced payments-much of it still manual or non-digital. That's inefficiency finance can convert into margin and retention.
See context on commercial payments opportunities
SMB automation is the wedge
SMBs face manual workloads, high labor costs, and error-prone processes. AI can read invoices, emails, and contracts, then automate AP/AR end to end.
Impact: lower operating cost, faster close, cleaner cash application, and better working-capital cycles.
Fintech playbook (what's working)
- Underwriting and merchant risk: lower loss rates, higher working-capital attachment
- Onboarding: intent-driven flows lift activation and cross-sell (PayPal cites 1.5x higher average margin per merchant and 2x more product activations)
- Fraud: AI models improving detection rates by 20% to 300%, blocking billions in attempts
- Embedded assistants: examples like Square's Managerbot streamline tasks and surface insights directly in merchant dashboards
What to do in the next 12-24 months
- Budget mix: prioritize data quality, access controls, model evaluation, and observability over "more models."
- Controlled pilots: run constrained, high-value use cases with clear guardrails (KYC/KYB, claims triage, agent assist, AP automation).
- Model strategy: mix foundational models with small, task-specific models; invest in retrieval for governed data access.
- Governance: formalize model risk policies, explainability standards, incident response, and human-in-the-loop checkpoints.
- Vendor posture: avoid lock-in; require offline evals, cost transparency, and clear data-use terms.
- Change management: rewrite workflows, not just swap tools; train frontline teams early.
- Regulator engagement: document assumptions, bias testing, and control design; keep audit trails tight.
Data readiness checklist
- Data inventory: know where sensitive fields live; standardize schemas and definitions
- Quality: set SLAs for completeness, accuracy, timeliness; monitor drift
- Access: granular entitlements; synthetic data where needed
- Lineage: track transformations; make them reviewable
- Privacy and security: encryption, tokenization, redaction, and strong secrets management
KPI scoreboard (measure what matters)
- Revenue: incremental lift per client, cross-sell rate, activation-to-adoption time
- Cost: cost-to-serve per case, minutes saved per workflow, unit cost per decision
- Risk: fraud basis points, false-positive/negative rates, model drift incidents
- Client experience: NPS/CSAT, first-contact resolution, time-to-refund/credit
- Control health: explainability coverage, audit pass rate, privacy incidents
Function-by-function quick wins
- Payments: dispute automation, real-time fraud scoring, chargeback reduction
- Retail banking: agent assist, personalized offers, intent-driven onboarding
- Wealth: next-best action, compliant content generation, meeting notes to CRM
- Insurance: claims intake triage, document extraction, subrogation detection
- Investment banking: deal screening, comps drafting, client memo generation (with strict review)
Talent plan
- Hire: AI compliance leads, model auditors, workflow/agent designers, data product managers
- Upskill: risk, ops, finance, and front-line teams on prompts, review standards, and exception handling
- Operating model: establish an AI program office with shared tooling and sandboxes serving multiple lines of business
The takeaway
AI's near-term story in finance is revenue expansion with higher opex. The firms that win will treat data as the product, build controls that scale, and train teams to work with machine collaborators.
Spend the next two years laying rails. Expect margin lift to show up as models stabilize and the data estate matures.
Further resources
Your membership also unlocks:
