September is only halfway through, but it’s already been one of the most exciting months of 2026 so far! On the back of our AI Summit in Singapore came the amazing news that we have been named a Leader in the inaugural Gartner® Magic Quadrant™ for Intelligent Document Processing (IDP) solutions. For me, this first-ever Magic Quadrant cements IDP as a must-have for modern enterprises. We believe this recognition reflects what our customers have experienced for years - ABBYY is the trusted choice for turning documents into business value with purpose-built AI. Get the full report here! But that’s not the only news we have for you this month. Read on to beat the odds that MIT recently presented when it comes to AI, Atari 2600s, digital twins, and what to bet on when it comes to Agentic AI. Like I said, exciting!
MIT Finds 95% of GenAI Pilots Fail - Here’s How to Beat the Odd
MIT's latest study highlights an uncomfortable truth: 95% of enterprise GenAI pilots fail to create business value. The problem, often called the “verification tax,” is that employees spend so much time checking GenAI’s work that efficiency gains evaporate.
But this points to a deeper issue. GenAI may generate convincing text or analysis, but enterprises struggle to embed it into day-to-day operations. Most pilots remain isolated experiments because what’s missing is a stronger foundation: a way to handle unstructured information reliably, monitor workflows, and enforce governance in real time. Without that, GenAI produces outputs — but not outcomes.
Structuring data and giving organizations visibility into their workflows, is what provides the grounding GenAI needs to scale. Rather than relying on constant human verification, technologies like Document AI and Process AI make it possible to apply GenAI with confidence — turning pilots into measurable impact.
The story of Gemini declining to challenge an Atari 2600 at chess might feel like a curiosity, but it reflects something more important: in AI, using the right tool for the right job matters. More scale or processing power doesn’t always deliver better outcomes. In fact, purpose-built systems often outperform general models because they’re designed with a specific task in mind.
Performance, too, should not be confused with intelligence. An LLM may generate convincing text, but even it will admit to “hallucinations” when pressed. That humility is actually a strength — a reminder that AI is still a tool, not an oracle.
The takeaway for enterprises is clear: success with AI doesn’t come from chasing the biggest model, but from thoughtfully orchestrating the right mix of specialized and general technologies. When matched to purpose, AI isn’t just powerful — it’s trustworthy, efficient, and transformative.
Salesforce’s own studies show that large language models alone achieve only 35% success rates in complex business scenarios.
To help bridge the gap between AI promise and performance, the company unveiled a “digital twin” of business operations where AI agents can be stress-tested before deployment.
Unlike existing benchmarks that test generic capabilities, the platform evaluates agents on real enterprise tasks like customer service escalations, sales forecasting, and supply chain disruptions using synthetic but realistic business data.
2025 is seen as the year AI agents make headlines, showing real value in software development, support, productivity, and customer engagement. However, enterprises require secure, reliable, and transparent solutions—not unpredictable experiments.
In his in-depth post, software engineer Utkarsh Kanwat examines the challenges of building AI agent-only solutions, including unsustainable expense at scale and compounding error rates.
Despite these challenges and difficulty keeping up with rapidly evolving AI technology, now’s the time to learn AI fundamentals. Then, review your existing workflows to see where it makes sense to refine them with AI. Start with a small, constrained, domain-specific use case that uses AI but maintains human control with strict boundaries over critical decisions. Finally, incorporate ABBYY’s end-to-end IDP capabilities to create a production-ready, reliable solution.
Before AutoGPT, There Was Shakey: The Original AI Agent
Back in 1960, researchers at MIT developed “Shakey the Robot”, often considered the world’s first true AI agent. Shakey could perceive its environment, plan actions, and execute them—essentially embodying the core loop of modern agentic AI systems. While today’s AI agents are digital, orchestrating tasks across software, Shakey rolled around on wheels, stacking blocks and navigating rooms. Fast forward to now, and the same principles underpin agentic systems coordinating across APIs, tools, and workflows in business.
Blog: ABBYY and IBM watsonx.ai Orchestrate Transform KYC Automation at Scale > By combining ABBYY’s advanced Document AI and Process AI with IBM watsonx.ai Orchestrate’s AI-driven agents, we’re transforming the entire Know Your Customer (KYC) process—from document intake to compliance monitoring—into a transparent, scalable and intelligent workflow. Read how.
Discover how process AI and document AI enable agentic systems to drive autonomy and business value.Read more (German).
ABBYY in Action: Innovation Spotlight
Tokyo Shoko Research Reduces Data Entry Time by 80% with ABBYY AI OCR
With a 133-year history, TSR is committed to quality and accuracy in its corporate credit investigations. That’s why TSR chose ABBYY to automate the processing of financial statements and real estate registration documents, at speed.
The Stock Market, an AI Report, and a 5% Success Rate
MIT researchers dropped a truth bomb that sent shockwaves through Wall Street: 95% of generative AI initiatives are failing to deliver measurable ROI. But here’s the plot twist that should keep C-suite executives up at night: it’s not the technology that’s broken. It’s how we’re wielding it.
October 6-9 – Orlando, FL - IBM TechXchange 2025 conference is a learning IBM flagship event that brings together developers, technologists, and techies looking to grow their skills. ABBYY will participate as an IBM integration partner with an exhibit space and a speaking session titled “Know Your Customer at Scale: ABBYY + IBM watsonx.Orchestrate Power an Agentic Approach to Onboarding.
October 30 - “From automation to autonomy - how process and document AI enable agentic transformation” - Join the powerful line-up of thought leaders from IDC and Lloyd’s Banking Group as they discuss this topic.
October 14 – ABBYY Ascend is our premier launch event series, designed to showcase how our latest innovations in Process and Document AI can drive real business impact. It’s a must attend for ABBYY customers and those who are new to ABBYY. Check out the recently added agenda.
State of Intelligent Automation: Generative AI Disillusionment
The rapid adoption of Generative AI (GenAI) is reshaping how businesses operate, but behind the hype, many leaders are grappling with unexpected challenges. Are companies truly reaping its potential, or are they fostering a shadow IT culture and adding to technical debt? Find out what decision makers really think about GenAI and learn what’s on their AI Wishlist.
September 23rd – 24th Manila Marriott Hotel, Philippines – ABBYY holds an Exhibitor pass at this event. 600+ pre-qualified Technology & Business Heads from all leading Banks, Insurance, Microfinance, Credit Unions, and Regulatory bodies across the country will be attending this event.
Nuix XLR8/25 - Nuix XLR8/25 - Sept 23rd Sydney Opera House, Sydney, Australia
NUIX XLR8/25 unites global experts and forward-thinking leaders to showcase how AI is amplifying human potential—not replacing it. Tailored for technically savvy NUIX customers like CTOs, CIOs, and Heads of Innovation, the event highlights real-world applications that drive transformative change. The ABBYY team will demonstrate cutting-edge OCR and ICR technologies that revolutionize document processing, enabling faster, smarter image-to-text conversion and recognition.
Discover the latest in Document AI, Process AI, and Generative AI through inspiring talks, live demos, and partner success stories. See how developers and partners are building impactful solutions with ABBYY technology—available anytime, anywhere. Begin with the highlight reel, then explore the full sessions from AI Pulse by ABBYY Developer Conference.
You have received this e-mail because you are an ABBYY customer or partner, have requested further information on our products, solutions or services, registered for an ABBYY trial version, and/or agreed to receive news from ABBYY.
