Editor’s Note

This article was adapted by EUVC from an original insight published by Synaptic.

Read the original article.

Financial services has used artificial intelligence for years, from algorithmic trading to risk assessment. Generative AI creates a different opportunity: applying large language models (LLMs) to the document-heavy, language-intensive work that still absorbs significant time across finance.

The first wave is not spreading evenly. Adoption is moving fastest where AI can extract, classify, summarise and retrieve information without directly determining a sensitive financial outcome. That makes back-office finance, compliance support, investment research and personal-finance interfaces natural early markets.

AI was already widespread in fintech before the generative-AI cycle. In 2020, nearly 90% of fintech firms surveyed by the Cambridge Centre for Alternative Finance said they used AI or machine learning to some degree.

LLMs add a new interface and a new automation layer. Finance teams handle transactions, compliance records, bills, invoices, reports and other unstructured documents. These models can extract key details, classify information, summarise records and assemble reports from multiple sources.

The productivity potential was already visible in Synaptic’s 2024 analysis. Brex said its expense-management AI assistant saved about 4,000 hours of manual work annually. Vic.ai reported that one large US distributor processed 76% of invoices without human intervention, averaging 36 seconds per invoice instead of 2.3 minutes.

The same capabilities can support customer service. Chatbots and knowledge assistants can interpret routine questions and retrieve answers, allowing finance professionals to focus on work that requires judgement.

Generative AI can support accounting, bookkeeping, transaction coding, expense tracking and the creation of receipts and memos. It can also detect patterns and exceptions, contribute to forecasting and surface personalised insights.

Basis, Arkifi and Vic.ai were among the companies applying a copilot model to finance teams. Sensible, Rossum and Attri focused on extracting information from annual reports, financial statements, earnings calls and other documents before moving the structured data into downstream systems.

Customer onboarding, anti-money-laundering checks and know-your-customer or know-your-business processes involve substantial documentation. Generative AI can assist with compliance reviews, identify missing documents and help draft responses to sanctions or suspicious-activity alerts.

Greenlite used generative AI to automate onboarding reviews. Coris used GPT-4 to help classify merchants for monitoring and compliance. Flagright applied GPT-4 to public sources such as Crunchbase and LinkedIn, and used it to draft Suspicious Activity Reports and routine analyst narratives.

AI interfaces can let investors query earnings releases, financial reports and market data in natural language. StockGPT focused on fundamental information for S&P 500 and Nasdaq companies. FinChat combined searchable financial data on more than 100,000 public companies with screening, visualisation, comparison and portfolio tools.

Consumer products used generative AI to make financial guidance more conversational and personalised. Wally summarised expenses and generated observations based on financial goals and habits. UStreet produced retirement plans, portfolio reviews and tax-saving guidance.

Financial services operates under demanding privacy, compliance and accuracy requirements. Synaptic therefore expected early adoption to concentrate in selective areas that do not directly determine major financial outcomes.

Expense tracking, document processing, investment research and personal-finance management can still involve sensitive information. But, as product categories, they can offer clearer paths to adoption than systems that directly approve credit, execute regulated advice or make other consequential decisions.

This constraint shapes both product design and market entry. Startups must prove that automation can save time without compromising review, auditability or data protection.

Fine-tuning models on relevant financial data is critical to creating useful finance workflows. That favours established firms with proprietary datasets and the infrastructure to train or adapt specialised systems.

BloombergGPT, JPMorgan’s IndexGPT, Intuit’s GenOS and Morgan Stanley’s AI assistant illustrated this incumbent advantage. These companies could combine domain data with established distribution and regulatory experience.

Companies with existing customer bases can integrate assistants into familiar workflows. Vise Intelligence and PilotGPT allowed users to query financial information in natural language, while Public.com’s Alpha delivered investing context and alerts.

New entrants concentrated on accounting, bookkeeping, expenses, underwriting, fraud, customer onboarding and compliance. Their common proposition was to remove routine work and give finance professionals more time for judgement-intensive tasks.

For additional context on how this progression extended from copilots to systems that execute workflows, see the related EUVC article on AI-agent use cases and adoption barriers.

Global fintech investment fell to a five-year low in 2023, but Synaptic reported that AI-focused fintech attracted more than $12.1bn during the year.

Among emerging AI-native companies, Synaptic counted more than 30 deals across 2022–24. Most were relatively small, reflecting cautious optimism: investors could fund product development and regulatory validation before committing larger sums.

Examples included Slope’s $30m round in 2023 after launching SlopeGPT and Federato’s $25m Series B for AI-supported underwriting.

The largest rounds went to established startups that had made generative AI central to their products:

The pattern supported Synaptic’s broader conclusion: investors were willing to make speculative bets on new tools, but the largest commitments went to companies with established products and visible market traction.

Synaptic’s web-traffic analysis showed interest moving from the initial AI surge towards more value-driven engagement. Aggregate traffic peaked in the first quarter of 2023 and then stabilised.

B2B products showed the strongest quarter-on-quarter growth in late 2023. Vise recorded growth of more than 500% in the fourth quarter, Kasisto grew 101% and Unit21 grew 77%. Consumer products such as Magnifi and FinChat still attracted substantial traffic, but growth was slowing as user behaviour shifted from curiosity towards practical utility.

These figures are a Q4 2023 snapshot, not current performance data. Their relevance lies in the pattern Synaptic identified: sustained engagement increasingly depended on a product solving a specific financial workflow.

Synaptic identified four early areas: CFO and finance tooling, onboarding and compliance, investment research, and personal finance.

They involve large volumes of documents and language-heavy tasks such as extraction, classification, summarisation and report generation.

Established firms control relevant proprietary data and already have the infrastructure and regulatory experience needed to build specialised models and workflows.

The central constraints are data quality, privacy, accuracy and regulation, especially when a system influences sensitive financial decisions.

Synaptic is a data and AI platform for private-market investors. It turns signals from financials, decks, websites, news and alternative metrics into connected data for sourcing, research and portfolio monitoring.

Learn more about Synaptic’s work in private-market research and portfolio intelligence.

Read the original article.

This article was adapted by EUVC from an original article published by Synaptic.

Read the original article.