Provectus

Most banks have an AI strategy. Far fewer have an AI risk strategy. That gap is now showing up in audits, regulatory reviews, and boardroom conversations – and it's getting harder to paper over. Helen Johnson (Executive Director, Financial Services at Provectus) and Evgueni Ivantsov (Chair, European Risk Management Council) just published a piece in The Banker on exactly this. Three things they argue CROs can no longer defer: 1️⃣ Model validation isn't a one-time exercise. AI models drift. They hallucinate. They behave inconsistently at the edges. If your sign-off process doesn't rerun when the model changes, the prompt changes, or the use case changes – it isn't a control. It's a formality. 2️⃣ AI ethics needs to be in the risk appetite statement, not the intranet. Where does your organization actually draw the line on customer profiling? On autonomous lending decisions? Those positions need to be explicit and tested. Vague commitments don't survive regulatory scrutiny. 3️⃣ AI concentration risk is the new cloud risk. Most firms can't tell you which processes would break if their primary AI vendor changed its terms tomorrow. That's the same exposure they once had with single cloud providers – and it took years to govern properly. This is the work we do at Provectus: helping financial organizations close the gap between AI ambition and AI readiness. 👉 Read full article: https://lnkd.in/dw_4kvXf What's the risk conversation you're not having yet in your organization?

The Anthropic CCAF exam is officially here! Our Director of AI Engineering and others on the team are now CCAF certified. Here are some tips from advanced practitioners at Provectus on preparing for the exam. To learn more on the topic, read the following Anthropic articles: ✔️ Context Engineering - https://lnkd.in/g3JrkuBK ✔️ Building Effective Agents - https://lnkd.in/eYHSwNvG ✔️ Effective Harnesses for Long-running Agents - https://lnkd.in/g24MEWgQ

AI moves fast. Our team moves faster. Provectus Academy is how we make that happen. Hear from our L&D Specialist Mariia Borysova on what it looks like to build a culture where learning is part of the job, not an afterthought.

AI agents write code faster than any human team. But can they write it securely? According to a 2025 benchmark from ETH Zurich, no model exceeded 37% correct and secure backend generation. And in DryRun Security's 2026 Agentic Coding Security Report, 87% of pull requests generated by Claude Code, OpenAI Codex, and Google Gemini contained at least one vulnerability. Not a single agent produced a fully secure application. But the problem isn't the agents, it's the security infrastructure built around human-paced development – and the fact that it hasn't kept up. Our latest technical guide by Konstantin Makarychev walks through a modernized shift-left architecture that embeds automated security scanning directly into AI coding workflows, before a vulnerable line ever hits a PR. Inside the guide: 🔸 Why conventional shift-left breaks down in agentic development 🔸 How Claude Code + Semgrep + MCP form a "scan-on-generate" architecture 🔸 Step-by-step setup: from local IDE to CI/CD pipeline 🔸 What the integration catches, where its limits are, and how to measure impact The goal: reduce Mean Time to Remediation (MTR) from days to minutes, so the efficiency gains from AI translate into quality, not technical debt. If your team is building with Claude Code (or planning to), this is the implementation guide you'll want in your toolkit. 👉 https://lnkd.in/eHPvKqme

Every company loses context. People leave, projects wrap up, decisions get buried in old threads. The knowledge stays somewhere in Slack, calls, and docs, but nobody can find it when it matters. That's why we built Barley, our GenAI assistant for corporate knowledge management. After 2 years of development, Barley pulls together data from Slack, Google Calendar, and our internal tools, like HOPS, to answer project questions, generate status reports, summarize meetings, and surface the context that would otherwise disappear. We talk a lot about enterprise-wide AI transformation for our clients. Barley is how we apply it to ourselves – in practice. Swipe through to hear from Vitalii Tafintcev, who has been leading Barley's development, on why we built our own solution and what it's changed.

Provectus keeps learning, and Q1 was another strong example of that. A big thank you to the team members who reached their certification goals this quarter. Your work, consistency, and commitment to growing your expertise matter. Here’s our Q1 Hall of Fame.

Thank you London! Provectus took home the Value Creation Award at last night's Actum Private Equity Value Creation Awards 2026 in the software category – a brilliant evening in the stunning National Gallery bringing together 300+ senior leaders from across the PE ecosystem . Our own Torsten Kriedt had the honor of accepting the award on behalf of Provectus, in recognition of our outstanding work delivering real, measurable AI-driven value creation for private equity and their portfolio companies. It's genuinely exciting to see AI value creation come of age. A big thank you to the team at Actum Group for putting on such a memorable evening, and to the jury for their time and rigor in recognizing excellence across the industry. Congratulations to all the winners of the night!

Legacy modernization doesn't have to start with code migration. In our latest article, Oleg Shuralev, Head of Application Development, shares a practical workflow for rebuilding legacy products with Anthropic's Claude Code by starting where the real value lives: in the product logic, user flows, and business knowledge embedded in the existing system. The key idea is simple. Extract the product first, define the future state clearly, and only then move into implementation with a structured, spec-driven workflow. Read the full article to see how Claude Code and AWOS can help modernize legacy applications with more speed, clarity, and control. 👉 https://lnkd.in/dhDQ7trW

February was a month of learning and development at Provectus. We explored topics like neurobiology, ran public speaking trainings, and launched our own Academy. Different formats, one goal. Help our teams grow in ways that matter for their day-to-day work. This is how we build real capability. Not through one-off sessions, but through continuous learning that turns into better thinking, better communication, and stronger delivery. And we’re just getting started.

Enterprise AI has reached an inflection point. We were in the room for it. The Provectus team was honored to take part in the Anthropic Inaugural Partner Kickoff in San Diego last week. The Anthropic team made it clear that they aim to become the AI tool of choice for enterprise use cases. Anthropic is leading from the front by building a partner program designed to invest in partners like Provectus. Our clients are already deploying Claude-powered agents in production, and the new partner framework allows us to move faster and go broader. This is the moment. We’re in it. Let's go 🚀 Colin Murphy, Oleg Blokhin, Helen Johnson, Matt Ewalt, Jerrod Hill – great work representing the team. #EnterpriseAI #AgenticAI #Anthropic #Provectus