What is Wise's AI strategy?
Wise has not published a standalone AI strategy or launched a flagship AI assistant; what it has said puts AI in engineering, customer support and financial-crime work. Its FY26 Form 20-F, filed June 25, 2026, says Wise has invested in AI and machine learning over the last decade, including for fighting financial crime, currency flow prediction and risk management, that machine-learning algorithms complete 99% of its compliance checks in less than one second, and that it has launched large language model copilots for its Financial Crime and Customer Support teams. In his letter to owners the same month, CEO Kristo Käärmann wrote that almost 50% of chat contacts are now handled by LLMs, "importantly, with improved quality", and that Wise's engineers use AI and now ship over 6,000 releases a month, up over 25% on the previous year. CTO and US president Harsh Sinha said in May 2026 that Wise had been careful in how it rolled out its LLM assistant because response quality matters. The one customer-facing AI feature, launched October 30, 2025, reads an uploaded invoice or screenshot and prefills payment details. The 20-F lists AI and machine learning as a risk factor, citing model errors in financial crime detection and transaction monitoring and possible scrutiny under the EU AI Act. Wise hosts its primary payment processing on AWS in Frankfurt; it does not disclose an AI budget, AI headcount or which model providers it uses.[1][2][3][4]
Data as of Oct 22, 2026 · Wise profile updated Oct 8, 2026 · 10 sources cited on this page · 52 fintechs profiled on this site · Markdown
What have Wise's leaders and filings said about AI, and when?
Statements and disclosures in date order. Wise's most detailed AI statements are in its Form 20-F and annual report rather than product launches.
| Statement or move | Who | When | Source |
|---|---|---|---|
| Launches 'Upload screenshot' / 'Upload invoice': generative AI prefills recipient details from a PDF or image; new customers save an average of 36 seconds per new recipient | Wise | October 30, 2025 | Wise blog[4] |
| Wise deployed an assistant that handles roughly 50% of customer queries via LLMs: "What you want is the tuning such that the quality of the responses are really good…so that's why we've been careful in how we roll this out" | Harsh Sinha, US president and CTO | May 2026 (reported May 22, 2026) | American Banker[3] |
| Invested in AI/ML over the last decade; processes large volumes of documents with AI-driven technology; launched LLM copilots for Financial Crime and Customer Support teams | Wise (Form 20-F) | June 25, 2026 | FY26 Form 20-F[1] |
| "Almost 50% of chat contacts are now handled by LLMs - importantly, with improved quality. We will continue to optimize our use of AI to achieve improved customer outcomes while being careful to maintain and improve customer satisfaction." | Kristo Käärmann, co-founder and CEO | June 2026 | Annual report letter to owners[2] |
| "Our engineers use AI with rapid improvements in productivity seen in the last year"; Wise ships over 6,000 releases a month, up over 25%, with over 1,000 engineers | Kristo Käärmann | June 2026 | Annual report letter to owners[2] |
| AI and ML listed as a risk factor: model errors could leave illicit activity undetected or block legitimate transactions; use cases may attract scrutiny under the EU AI Act; GDPR Article 22 cited | Wise (Form 20-F) | June 25, 2026 | FY26 Form 20-F[1] |
| Technology costs include company-wide tools "including AI solutions", and servicing costs include "technology and AI solutions used by servicing teams"; Wise expensed $226.8 million of product engineering costs in FY26 (FY25: $164.6 million) without breaking out AI | Wise (Form 20-F) | June 25, 2026 | FY26 Form 20-F[1] |
| Q2 FY27 results statement covers volumes, new markets and Wise Platform partners; it does not mention AI | Kristo Käärmann | October 15, 2026 | Q2 FY27 results[5] |
Where does Wise use AI and machine learning?
Wise rarely dates its internal AI tools; the 'Since' column gives the date of the first disclosure.
| Product | What it does | Built on | Since |
|---|---|---|---|
| Compliance checks | 99% of compliance checks completed in less than one second | Machine-learning algorithms | Disclosed in the FY26 Form 20-F (June 25, 2026)[1] |
| Real-time risk alerts | Alerts that help global teams detect and mitigate risks | ML models trained on data points and patterns across millions of customer profiles and their transactions | Disclosed June 25, 2026[1] |
| LLM copilots | Help Financial Crime and Customer Support agents with investigative precision and efficiency | Large language models; providers not named | Disclosed June 25, 2026[1] |
| LLM customer support chat | Handles almost 50% of chat contacts, per the CEO; roughly 50% of customer queries, per the CTO | Large language models; providers not named | May 2026[2][3] |
| Document processing and currency flow prediction | Processes large volumes of documents; predicts currency flows; supports risk management | AI-driven technology; AI/ML models | Disclosed June 25, 2026[1] |
| Upload screenshot / Upload invoice | Reads an uploaded invoice, bill or screenshot and prefills recipient details and amount; customers review before sending | Generative AI | October 30, 2025[4] |
| AI for engineering | Engineers use AI; over 6,000 releases a month | Tools not named | June 2026[2] |
| Infrastructure | Primary payment processing; Amazon Web Services is named as a critical third-party provider | AWS Frankfurt region | Disclosed June 25, 2026[1] |
Who runs AI at Wise, why is it cautious, and what does it not disclose?
Wise has no chief AI officer on the record. Technology sits with Harsh Sinha, CTO since May 2015, who American Banker identified in May 2026 as US president as well; product with chief product officer Nilan Peiris. The 20-F says over 1,000 engineers across four continents run a platform processing over 4.7 million transactions a day.[1][3]
AI has not shrunk the service workforce: servicing staff rose to 5,580 at March 31, 2026 from 4,136 a year earlier, and Wise said on June 1, 2026 that around one third of its global team is dedicated to protecting customers from financial crime.[1][6]
The caution has a regulatory backdrop. The machine learning Wise uses in transaction monitoring sits in the area where a 2025 US multistate consent order found data-integrity concerns, cited by the OCC when it denied Wise's trust-bank charter in July 2026. On its October 15, 2026 earnings call, CFO Emmanuel Thomassin said Wise had "provided the full remediation on every point that was part of the consent order" and was waiting for feedback from the authorities.[7][8]
What is not disclosed: an AI budget or AI headcount, which model providers power the LLM copilots and support chat, quality or accuracy metrics for those systems, and any plans for AI agents that act for customers. Wise has made no statement on agentic payments.[1][2][5]
What does the Wise profile record on this topic?
The rows below are the entries of the Wise AI strategy profile that bear on this page, as of Oct 8, 2026; the profile holds the full record of 10 dated moves, 7 use cases and 12 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| LLM copilots for Financial Crime and Customer Support | Wise has launched large language model copilots across teams, including Financial Crime and Customer Support, to help agents improve investigative precision and efficiency. | In production | [1] |
| Upload screenshot / Upload invoice | Generative AI reads an uploaded PDF or image and prefills recipient details and amount; new customers save an average of 36 seconds per new recipient, according to Wise. | In production | [4] |
| ML currency flow prediction | Wise uses AI/ML for currency flow prediction and risk management, and its 20-F says errors or biased data in these models could produce inaccurate outputs, regulatory action or financial loss. | In production | [1] |
| Third-party generative AI systems | The 20-F says Wise uses third-party generative AI systems and flags risks from training-data copyright and from staff or vendors entering confidential information into third-party AI tools. | In production | [1] |
| Recommended by AI shopping/banking assistants | Independent research found large language models (ChatGPT, Gemini, Perplexity) recommend Wise, alongside Mercury and Bluevine, to small-business owners more often than traditional banks when asked for a specific account recommendation. | In production | [9] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Sep 24, 2026 | Study: AI chatbots recommend Wise over traditional banks | American Banker reported on a study by research firm AIVO in which researchers posing as small-business owners had 2,160 conversations with ChatGPT, Gemini and Perplexity; across the three models, fintechs (including Wise, Mercury and Bluevine) drew 1,345 recommendations versus 452 for traditional banks when the AI was asked for a specific product recommendation. | [9] |
| Jun 25, 2026 | FY26 results and first Form 20-F detail AI use | Wise reported net revenue of $2,502.8 million (up 19%) and 18.9 million active customers; its 20-F disclosed LLM copilots for Financial Crime and Customer Support teams and ML-driven compliance checks, and listed AI model risk among its risk factors. | [10][1] |
| Oct 30, 2025 | Wise launches AI 'Upload screenshot' for payments | A generative-AI feature reads an uploaded invoice, bill or screenshot and prefills recipient details and transfer amount; available globally to personal and business customers, who must review the data before sending. | [4] |
What is Wise's AI strategy?
Wise has not published one as such. Its filings and CEO letter describe AI used behind the scenes: machine learning for compliance checks and risk alerts, large language model copilots for financial-crime and support staff, LLMs that handle almost half of support chats, and AI tools for engineers, alongside one customer feature that reads invoices.
Does Wise use AI in customer support?
Yes. CEO Kristo Käärmann wrote in June 2026 that almost 50% of chat contacts are now handled by LLMs with improved quality, and CTO Harsh Sinha said in May 2026 that an assistant handles roughly 50% of customer queries. Wise's 20-F also describes LLM copilots for its Customer Support team.
Who leads technology and AI at Wise?
Harsh Sinha has been chief technology officer since May 2015 and was described by American Banker in May 2026 as Wise's US president as well. Wise has not named a chief AI officer. Co-founder Kristo Käärmann is CEO and Nilan Peiris is chief product officer.
What AI risks does Wise disclose?
Its FY26 Form 20-F says errors or biased data in the AI and ML models it uses for financial crime detection, transaction monitoring and currency flow prediction could miss illicit activity or block legitimate payments, that its AI use cases may attract EU AI Act scrutiny, and that staff or vendors could leak data into third-party AI tools.
Which AI models or cloud does Wise use?
Wise has not named its model providers; its 20-F says it uses third-party generative AI systems. The filing says its primary payment processing is hosted in Amazon Web Services' Frankfurt region and names AWS as a critical third-party service provider.
- Wise Group plc Annual Report on Form 20-F for the fiscal year ended March 31, 2026
- Dear Wise Owners: FY2026 letter to owners (Exhibit 99.1 to Form 6-K)
- AI beats humans at transaction monitoring, Revolut US CEO says
- Spend Less Time Typing: The Faster Way to Pay Invoices with Wise
- Wise Group plc reports Q2 FY27 Results
- Wise's response to recent press articles (Form 6-K)
- Corporate Decision #1381: Application to Charter a Subsidiary National Trust Bank, Wise National Trust (denied)
- Wise on AML: 'It's difficult to get everything right'
- Why AI models favor fintechs over traditional banks
- Wise Group plc reports Full Year 2026 Financial Results (Exhibit 99.1 to Form 6-K)
When Wise moves on AI, you'll read it here first.
when a bank, fintech or regulator 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