82% of mortgage brokers use ChatGPT, but readiness and marketing adoption lag

82% of Brokers Are Using ChatGPT - But Most Aren't Ready to Scale It

AI is already on the desks of mortgage brokers. A new Paradigm survey shows nearly 82% have used ChatGPT or similar tools in the last three months, and 87% want to learn more.

The gap: fewer than 30% are using AI for marketing, and confidence is low across technical areas. Only 45% say they understand the difference between open and closed data. When asked to rate their AI alignment (1-5) across data governance, workforce readiness, and an actionable roadmap, scores averaged below 3.

Key findings at a glance

• 82% have used AI tools recently.
• 87% want deeper AI education.
• <30% are using AI in marketing.
• 45% feel clear on open vs closed data.
• Sub-3 average scores on governance, readiness, and roadmaps.

What industry leaders are saying

Paradigm's mortgage services director Richard Howes says brokers are leaning in: "They see AI as something to walk towards rather than away from." He adds that upcoming guidance will help firms "market their business using AI and large language models," while defending against "bad actors who will inevitably seek to exploit this technology."

OSB chief credit and MLRO Richard Wilson underscores the risk side: "Authenticating the information you receive is crucial, as is ensuring the techniques you use to summarise or explain data are safe, trusted, and understood."

Why this matters for finance teams

Margin pressure, tighter oversight, and higher client expectations mean productivity gains matter. AI can reduce turnaround times, sharpen risk reviews, and deliver cleaner client comms - if you set guardrails first.

The weak link is data. Think in two buckets: open data (public, non-sensitive) and closed data (client, commercial, or regulated). Mixing them in the wrong tools creates leakage, bias, and audit issues. Clear rules, private deployments, and zero-retention modes are your first lines of defense. For regulatory context, see the ICO's guidance on AI and data protection here.

Practical wins you can ship this quarter

• Marketing (low risk): Draft emails, social posts, landing-page copy, and FAQs. Keep compliance review in the loop. Log prompts and outputs.
• Operations: Summarise case notes, extract key facts from documents, build checklists, and standardise handovers. Use templated prompts for consistency.
• Client service: Generate plain-English explanations of products, fees, and timelines. Add disclaimers and link to source documents.
• Quality and audit: Auto-flag missing documents, inconsistent data, and potential AML red flags for human review.

Minimum viable AI policy (start here)

• Data rules: No client or confidential data in public models. Use enterprise/zero-retention settings. Mask PII where possible.
• Model access: Approved tools list, SSO enforced, audit logs on, role-based permissions.
• Prompt hygiene: Template library, source citation required, no advice without human sign-off.
• Validation: Human review for outputs that affect clients, credit decisions, or regulatory submissions.
• Vendor checks: DPAs in place, data residency documented, security attestations (SOC 2/ISO 27001), AML/financial crime implications assessed.

Reduce fraud and misinformation risk

• Verify sources: Require links to origin documents for summaries. Prefer internal repositories or CMS over the open web.
• Content authenticity: Use watermarks and version control for client-facing materials. Keep an audit trail.
• Staff training: Spot synthetic IDs, spoofed docs, and AI-written phishing. Align with AML and KYC procedures.

A 30-day plan to get from testing to traction

• Week 1: Approve 1-2 tools, publish a one-page policy, set up audit logging. Pick three safe use cases.
• Week 2: Build prompt templates. Run a workshop. Define what must be human-reviewed.
• Week 3: Pilot with 5-10 users. Measure time saved, error rates, and compliance feedback.
• Week 4: Kill what didn't work, standardise what did, and write a simple 90-day roadmap.

Training and next steps

Paradigm plans to work with lenders and brokers to roll out practical training so firms can adopt AI responsibly and profitably. That's the right focus: start small, keep data safe, and build repeatable workflows.

If you want curated tools specific to finance, this catalog is a useful shortcut: AI tools for finance.

Bottom line

Adoption is high, but maturity is low. The firms that win will pair simple use cases with tight controls, clear metrics, and steady training. Do less, better - then scale what works.