What proprietary AI has Flagstar Bank built in-house?
Flagstar's in-house AI is one system, StarIQ, and it is an orchestration layer rather than a model of its own. Deployed in early 2026, StarIQ routes work to foundation models from Anthropic (Claude), Meta (Llama), Mistral and Amazon (Titan), runs on Amazon Web Services and is secured by Palo Alto Networks; what Flagstar built is the layer around them: consumption-based multi-model orchestration without per-seat licences, an AI-aware security layer, a custom retrieval pipeline with citation tracking, and a governance workflow from business case to multi-level approval. On June 15, 2026 the bank said it had filed a provisional patent application for it, 'Techniques for Secure Enterprise Generative Artificial Intelligence Orchestration'. Chief financial officer Lee Smith said in May 2026 that StarIQ is available to all 5,400 employees and that about 83% to 84% use it regularly. In July 2026 chief information and operations officer Chris Higgins said it sends each request to one of three model tiers, runs in an AWS tenant, was built by his distinguished engineers, and is used most by risk management and marketing. No customer-facing AI, model trained by Flagstar, or StarIQ cost figure is on the record.[1][2][3][4]
Data as of Oct 22, 2026 · Flagstar profile updated Oct 13, 2026 · 11 sources cited on this page · 140 banks profiled on this site · Markdown
Which AI systems has Flagstar built, and what do they run on?
'Built on' names the models and infrastructure as the bank described them. 'Since' is the earliest public date.
| Platform or model | What it does | Built on | Since | Status |
|---|---|---|---|---|
| StarIQ orchestration layer | Integrates several AI providers in one governed platform under consumption-based deployment, 'eliminating per-seat limitations' | Anthropic Claude, Meta Llama, Mistral and Amazon Titan on Amazon Web Services | Early 2026 | In production; provisional patent filed June 15, 2026[1] |
| Model tiering | Three tiers Higgins likens to a bachelor's, master's and PhD degree; requests do not default to the top tier, and engineered criteria decide which tier a task gets | Foundation models inside StarIQ; tier-to-model mapping not disclosed | July 2026 (on the record) | In production[3] |
| AI-aware security layer | Real-time detection of sensitive data patterns, prompt injection attempts and policy violations | Palo Alto Networks Prisma AI firewalls | Early 2026 | In production[1] |
| Custom RAG pipeline | Document and policy queries with full citation tracking and access-controlled knowledge bases | Flagstar's own documents; built by the bank | Early 2026 | In production[1] |
| Branch and call-centre assist | StarIQ is trained on compliance and regulation, with branch and call-centre procedures and policies loaded, so tellers, bankers and agents get faster, correct answers | StarIQ retrieval over internal content | July 2026 (on the record) | In production[3] |
| Employee workspace | Analyzes company records, policies and procedures and helps with presentations and marketing materials; open to all 5,400 employees | StarIQ | May 2026 (on the record) | About 83% to 84% of employees use it regularly[2] |
| Integrated governance workflow | Takes each AI use case from business case submission through regulatory compliance review and multi-level approvals in one auditable system | Built by the bank | Early 2026 | In production[1] |
| Flagstar S2 Platform | The enterprise platform StarIQ belongs to, one of its 12 components; trademark application filed | Bank-designed; eight outside firms work on S2 | June 2026 (launch) | In production[3][1][5] |
Which people and teams build Flagstar's in-house AI?
| Person or team | Role | Since |
|---|---|---|
| Chris Higgins, Chief Information and Operations Officer | Joined to recruit 'an exceptional technology team to, honestly, rebuild the bank'; chairs the AI strategy committee, which the CEO attends | October 2024[4][3] |
| Jason Pope, Chief Technology Officer | Frames the patent as part of 'building, not just buying technology'; the five March 2026 hires report to him | On the record from March 2026[1][6] |
| Distinguished engineers | Higgins: 'my distinguished engineers, very nice cadre, they're the ones who built it' | Before July 2026[3] |
| Robert Martin, Head of AI and Engineering | SVP and distinguished engineer; leads incubation across the Chief Technology Office; set up AI risk governance frameworks at MUFG | March 16, 2026[6] |
| Justin Zimmerman, Distinguished Engineer, Cloud Services | Led governance and AI-driven compliance work across AWS and Azure at U.S. Bank; brings AI-enabled governance capabilities to Flagstar's cloud platform | March 16, 2026[6] |
| Eric Gunn, Distinguished Engineer, Information Security | Applies machine learning to cybersecurity; a named inventor on multiple U.S. patents | March 16, 2026[6] |
| Enterprise Technology & Operations team with Cognizant | Cognizant works alongside the in-house team on platform development and AI-driven initiatives, under a hybrid-managed services model; the two plan to explore an AI Center of Excellence in India | 2025-2026[7][8] |
Why did Flagstar build StarIQ instead of buying an AI platform?
The June 2026 announcement says that 'rather than relying on third-party solutions', Flagstar architected a technology foundation of its own. Pope called the patent filing proof of 'building, not just buying technology, that is purpose-built for regulated financial services'.[1]
Cost control is the reason Higgins gives most often. 'That's actually why we built our proprietary solutions, StarIQ, not relying on tokenization,' he said in July 2026, adding that he would not put a foundation model accessed by API into a highly scaled process because that introduces variable expense, and that many tasks do not need 'a model that knows advanced calculus'.[3]
StarIQ is where experimentation happens; when a business case is chosen, Flagstar then builds for scale outside it, limiting itself to 'four, five, six meaningful implementations' at a time. Higgins said one of the leading principles when StarIQ was designed was 'high audit ability', so the bank can produce step-by-step detail on every transformation.[3]
The models themselves are bought. Banking Dive reported that StarIQ 'leans on multiple large language models including Anthropic's Claude, Meta's Llama and French company Mistral', and Higgins said more lenders may build their own platforms 'if they can recruit the talent'.[4]
What has Flagstar not disclosed about StarIQ?
The patent application is provisional, so its claims are not public. Flagstar has not said which model sits in which of StarIQ's three tiers, how many requests StarIQ handles, or how many engineers built it.[1][3]
Higgins declined to give the size of the investment in S2 and StarIQ, calling it 'significant for a company of our size'; first-quarter 2026 software expense rose 12% to $47 million. A $90 million technology and operations modernization effort was approved in June 2025.[4][7]
No customer-facing AI is in production: Higgins said in July 2026 that AI at the bank is 'mostly behind the scenes' and that customers are probably six to nine months from noticing it. The 2025 10-K contains no AI risk factor.[3][9]
What does the Flagstar profile record on this topic?
The rows below are the entries of the Flagstar AI strategy profile that bear on this page, as of Oct 13, 2026; the profile holds the full record of 29 dated moves, 14 use cases and 30 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| StarIQ multi-model orchestration | Claude, Llama, Mistral and Titan behind one governed platform with consumption-based deployment. | In production | [1] |
| RAG over documents and policies | Secure queries with citation tracking and access-controlled knowledge bases. | In production | [1] |
| StarIQ assist for branch and call-centre staff | Trained on compliance, regulation and branch and call-centre policies and procedures so tellers, bankers and agents get faster, correct answers. | In production | [3] |
| StarIQ for underwriting data discovery | Credit underwriting uses StarIQ to find the data most relevant to faster decisions; risk management and marketing are the heaviest users. | Pilot | [3] |
| AI strategy committee and tiered model routing | The CIOO chairs an AI strategy committee attended by the CEO that prioritises a revenue and expense-reduction backlog; StarIQ routes work to three model tiers by complexity to manage cost. | In production | [3] |
| Machine learning in security engineering | Eric Gunn, hired as Distinguished Engineer for Information Security, applies machine learning to cybersecurity problems; Flagstar has not described specific deployed tools beyond the StarIQ firewall layer. | Announced | [6] |
| Head of AI with AI risk governance background | Robert Martin, who established AI risk governance frameworks at MUFG, now strengthens governance frameworks and least-privilege access as Head of AI and Engineering; the board Risk Committee hears from the CRO, CCO and CISO. | In production | [6][10] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Jul 30, 2026 | Inside StarIQ: who uses it and how it is governed | Higgins on American Banker's Leaders: one of 12 S2 components, trained on compliance and branch and call-centre procedures, three model tiers to control cost, runs in an AWS tenant, built for auditability; heaviest users are risk management and marketing; customers should notice in six to nine months. | [3] |
| Jun 15, 2026 | S2 trademark filed; StarIQ patent application | The bank establishes proprietary rights over its platform and its gen-AI orchestration system. | [1][11] |
| Mar 16, 2026 | Robert Martin appointed Head of AI and Engineering | One of the five March hires; he came from MUFG, where he built AI risk governance frameworks, and leads incubation in the Chief Technology Office plus identity, cloud, SRE and platform work. Eric Gunn (security, machine learning) and Justin Zimmerman (AI-driven compliance in cloud) joined the same day. | [6] |
| Jan 29, 2026 | Cognizant partnership marks a year; AI centre of excellence to be explored | Cognizant works as an extension of the in-house technology team on S2, with co-branded development centres in Hyderabad and Manila; next, the two plan to explore an AI Center of Excellence in India. | [8] |
Does Flagstar build its own AI models?
No. Flagstar built StarIQ, an orchestration, security, retrieval and governance layer, but the foundation models inside it are bought: Anthropic Claude, Meta Llama, Mistral and Amazon Titan, running on Amazon Web Services. The bank filed a provisional patent application for StarIQ on June 15, 2026.
How many Flagstar employees use StarIQ?
Chief financial officer Lee Smith said in May 2026 that StarIQ is available to all 5,400 employees and that about 83% to 84% use it regularly, to analyze company records, policies and procedures and to help with presentations and marketing materials.
Who uses StarIQ the most at Flagstar?
Risk management and marketing, according to chief information and operations officer Chris Higgins in July 2026. He said he had expected the technology organization to lead usage; credit underwriting is also using it to find data for faster decisions.
How does StarIQ control AI costs?
StarIQ sends each task to one of three model tiers that Higgins compares to a bachelor's, master's and PhD degree, and it does not default to the most capable tier. Higgins says consumption-based deployment without per-seat licences was a reason to build rather than buy.
Who built StarIQ?
Higgins credits his distinguished engineers. The technology organization under CTO Jason Pope added five senior hires in March 2026, including Robert Martin as Head of AI and Engineering, and works with Cognizant as an extension of the in-house team.
- Flagstar Bank Announces Proprietary Technology Transformation
- Flagstar Bank Touts Turnaround Progress, Eyes Capital Returns as CRE Exposure Falls
- The AI-Enabled Bank: Reimagining Retail Banking from Branch to Mobile (Leaders, with Chris Higgins)
- Flagstar CIO: Proprietary tech creates 'optionality'
- How Flagstar Bank integrated its acquired data centers
- Flagstar Bank, N.A. Expands Technology Leadership Team Accelerating the Bank's Technology Architecture
- Flagstar Bank's Technology Expansion: A Step Toward Strategic Growth
- Cognizant and Flagstar Bank, N.A. Mark One-Year Partnership, Advancing Flagstar's Transformation, Enhancing Client Experience and Driving Efficiency
- Flagstar Bank, N.A. Annual Report on Form 10-K for the fiscal year ended December 31, 2025
- Flagstar Bank, N.A. Proxy Statement for the 2026 Annual Meeting (Form DEF 14A)
- Flagstar Bank Announces Proprietary Technology Transformation (wire)
When Flagstar moves on AI, you'll read it here first.
when one of the banks we cover moves on AI, it is in the next morning's brief · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning