As generative AI sweeps across the corporate landscape, few sectors face more pressure—and more risk—than public markets. Investor relations teams are expected to deliver instant, verifiable answers, yet the rise of general-purpose chatbots has introduced a new challenge: speed without accuracy can quickly become a liability.
Vancouver-based Versance.ai is tackling that problem head-on. Founded with a compliance-first philosophy and built specifically for the regulated world of public issuers, the company has designed what CEO George Fleming calls a “reasoning system,” not another chatbot.
In a recent conversation with Techcouver, Fleming explained how Versance rethinks retrieval, auditability, and disclosure integrity to help public companies serve investors with confidence in an era where trust—and truth—are tightly scrutinized.
As CEO of Versance.ai, can you tell us in simple terms how Versance.ai is different from the typical chatbot that many people associate with AI today?
GF: Firstly, Versance isn’t a chatbot but rather a compliance-grade reasoning system for public companies. Instead of paraphrasing whatever it finds, our agentic AI searches a company’s official records (filings, releases, decks, website, earnings calls); checks freshness and provenance; and, answers with citations and dates. Our accuracy-first, compliance-first design lets IR and legal audit responses in seconds and trust them for investor-facing use.
You say your platform was built specifically for public-market environments. Why does that matter?
GF: It matters because public markets are a regulated truth environment. Every statement has to match the official record, and when it was disclosed often matters as much as what was said. Versance is engineered for verification over generation. Everything the system says can be traced back to a source document and this accountability is essential in capital markets.
Can you walk us through the architecture and core design of the platform?
GF: Absolutely. Versance.ai is built on a three-layer architecture designed for auditability:
Global Data Foundation: We process issuer filings, news releases, and disclosures across multiple jurisdictions to create a normalized, always-current dataset.
Securities-Compliant AI Engine: Our reasoning layer is built with regulatory guardrails fcovering requirements like Reg FD and NI 51-102. Every response links to its originating source.
Multi-Issuer Application Suite: This includes our IR Agent, Investor Research Assistant, Public Company Co-Pilot, SEDAR/EDGAR research tools, compliance monitoring, and AI-assisted content creation. Each application is powered by the same disciplined, evidence-first approach.
What does “agentic retrieval and reasoning” mean for the user?
GF: Agentic retrieval and reasoning means the system does the hard thinking for you, not to you. When you ask a question, it plans a search, tries multiple retrieval paths (semantic, keyword, temporal), and reformulates the query if needed. It then assembles time-aware, source-cited evidence from the company’s official record and only then composes an answer or flags uncertainty instead of guessing. For the user, the result is simple: trusted answers with the receipts, and behavior that stays consistent because every change is gated by permanent evaluation scenarios before it reaches production.
Answering investor questions at scale can be compute-intensive. How does Versance keep AI costs predictable, and what’s your pricing model?
GF: We designed the platform to be accurate and efficient. Our agentic retrieval does the heavy lifting up front – routing to the cheapest capable model, reformulating queries, and pulling only the minimal, verified context from a pre-indexed company corpus. We cache results, dedupe sources, and reuse citations, which cuts tokens dramatically versus ‘chatbot summarizes the internet’ approaches. In practice that means lower, steadier compute, even when volumes spike.
Pricing is simple and transparent: a flat monthly platform fee (compliance-grade search, apps, analytics) and, if a customer exceeds their included usage, LLM overage at our cost – no markup. As model prices fall and performance rises, we pass those savings through and raise the bar on quality. The goal is predictable OPEX for issuers, with auditable answers that teams can trust.
From a business perspective, why should a public company adopt Versance.ai?
GF: Look at the market reality: ChatGPT is touching ~800 million weekly users today – AI is in everyone’s workflow, and that includes a huge share of retail investors. So the bar has moved: investors expect instant, verifiable answers 24/7.
Beyond the obvious reasons for deploying AI there are four major benefits that Versance delivers:
Trust and Transparency: Investors can see exactly where an answer came from: the filing, the date, the line item. That builds confidence.
Regulatory Reliability: Because everything is evidentiary, you reduce the risk of inconsistent or non-compliant communication.
Operational Efficiency: IR teams spend less time handling repetitive questions and more time on strategic engagement.
Enhanced Investor Experience: Investors now expect fast, high-quality digital engagement. Versance delivers that while maintaining compliance-grade accuracy.We evaluate responses across factuality, completeness, relevance, tone compliance, and evidence traceability.
AI hallucinations are a major concern. How does Versance mitigate that risk?
GF: We treat hallucinations as a governance problem, not just a model problem. Four layers keep us honest:
Scope control. Answers are confined to the issuer’s approved corpus; the system can’t wander. Open-web retrieval is off by default and whitelisted when used.
Evidence thresholds & refusal policy. We require sufficient, time-consistent evidence to answer. If sources conflict or are thin, the agent declines rather than speculate.
Change management. Any tweak to models, prompts, or routing must pass a permanent evaluation suite before release, with canary deploys and instant rollback if behavior shifts.
Operational auditability. Every response carries citations, dates, and an audit trail. We monitor live metrics like refusal accuracy, unresolved-evidence rate, and point-in-time mismatches, and we sample outputs weekly for IR/legal review.
Net net: we don’t just hope a model won’t hallucinate; we design the system and the operating procedures so that when uncertainty shows up, it’s caught, surfaced, or stopped before it reaches investors
Looking ahead, what role do you hope Versance.ai will play in the future of capital markets?
GF: AI is becoming the default interface between investors and information. Public companies that lean in turn transparency into a competitive edge – delivering a premium investor experience with immediate, source-cited, auditable answers and personalized 24/7 engagement. Versance’s role is to provide an issuer-controlled, compliance-first AI platform that turns the official record into verifiable intelligence and puts it to work across investor Q&A, research, and disclosure.
Versance becomes a trust multiplier – investors get faster, clearer access; issuers reduce friction and ultimately support multiple expansion and a lower cost of capital.
How does a public company get started with Versance.ai?
GF: Visit Versance.ai and book a demo. Every listed issuer in Canada and the US is pre-loaded on our platform with its filings and news releases, so getting started is fast: we show you real deployments, switch on your issuer workspace, expand your knowledge base to include decks, site content, earnings-call transcripts, and optimize the IR Agent on your record. We run a short pilot, provide you with accuracy testing results that show 99% answer appropriateness – you approve and launch is as simple as adding a widget to your site. No heavy IT; Versance is compliance-first, source-cited from day one.
Leave a Reply