This Month in AI - September 2025

Welcome to the September 2025 edition of "This Month in AI." Here you'll find an overview of the latest articles and reports on advancements in artificial intelligence, trends in enterprise adoption, and innovative strategies business leaders like you are using to harness AI's transformative power.

Here’s what’s making news this month:

Crossing the GenAI Divide: Why Partnerships Win Twice as Often

It’s the billion-dollar question every boardroom is asking: if nearly 95% of enterprise AI projects fail, why are a few companies racing ahead with measurable ROI and double-digit growth? According to MIT’s The GenAI Divide: State of AI in Business 2025, the secret lies in leveraging systems that learn, remember, and adapt, alongside a strategic focus on leveraging proven third-party technology rather than internal development.

‍Research Highlights:

• 95% of initiatives fail when companies rely on static tools that don’t integrate with workflows or evolve with use.
• Projects leveraging proven third-party technology succeed 2x more often than internal builds, driving faster deployments, higher adoption, and lower cost.
• Employees are already ahead: 40% of knowledge workers use ChatGPT and similar tools personally, but abandon rigid enterprise versions that lack memory or persistence.
• Agentic AI shows promise: systems with persistent memory and autonomy are proving effective in customer service, financial processing, and sales.
• Winning strategies: Startups and enterprises crossing the divide share three traits:
  • Focus on narrow, high-value use cases.
  • Embed deeply into workflows.
  • Scale through continuous learning

The takeaway:‍

Crossing the GenAI Divide isn’t about building everything in-house. The organizations that win are partnering strategically, embedding adaptive systems, and letting users lead the way through minimal human intervention.

‍

Google Cloud’s ROI of AI Study Shows Agentic AI Early Adopters Pulling Ahead

Google Cloud’s latest ROI study surveyed 3,466 senior leaders of global enterprises across 24 countries with generative AI deployment within their organizations. The study finds that over half of enterprises already use AI agents, with early adopters reporting 6–10% revenue growth.

​The conversation has moved from 'if' to 'how fast,' and the new differentiator is agentic AI. Early adopters of agents are not just automating tasks; they are also redesigning core business processes. By championing AI as a core engine for competitive growth and thus securing dedicated budgets, they provide a clear roadmap for any organization looking to scale, solve complex challenges, and achieve more consistent ROI.

- Oliver Parker, VP, Global Generative AI Go-To-Market, Google Cloud.

‍

• Wider adoption: 52% of executives say their organizations now use AI agents, with 39% deploying more than ten.
• Early adopter advantage: 13% of leaders surveyed dedicate at least half their AI budgets to agents. Among them, 88% see ROI in at least one use case, vs. 74% average.
• Business impact: Agentic AI early adopters report stronger returns in customer service (43% vs. 36%), marketing (41% vs. 33%), security operations (40% vs. 30%), and software development (37% vs. 27%).
• Revenue growth: Over half (53%) of executives reporting gains cite 6–10% revenue growth from generative AI, consistent year-over-year.
• Industry and regional patterns: Top use cases include customer experience, marketing, security, and tech support, with financial services, retail/CPG, and telecom highlighting industry-specific deployments. Europe prioritizes tech support, JAPAC emphasizes customer service, and Latin America focuses on marketing use cases.
• Shifting concerns: As investment rises (77% report increased spending), privacy and security now top the list of considerations for choosing LLM providers, followed by ease of integration and cost considerations.
  ‍

AI Agents are the Next Enterprise Standard

By the end of 2025, most enterprise apps will embed AI assistants, setting the stage for agent-driven ecosystems, according to Gartner. Executives have just three to six months to set their agentic AI strategies or risk losing ground. That growth is expected to accelerate quickly: fewer than 5% of enterprise applications today have task-specific AI agents, but adoption is projected to surge to 40% by 2026. Longer-term, agentic AI could reshape the market, driving more than $450 billion in software revenue by 2035—about 30% of the total, up from just 2% in 2025.

‍

Five stages of evolution:‍

• Assistants everywhere (2025): Most enterprise apps embed assistants, though “agent washing” blurs definitions.
• Task-specific agents (2026): AI handles complex, end-to-end tasks like cybersecurity response.
• Collaborative agents (2027): Teams of agents work together within apps.
• Agent ecosystems (2028): Agents collaborate across apps and business functions.
• Democratized apps (2029): Half of knowledge workers gain skills to govern and build agents.

Enterprise apps will shift from static tools to dynamic agentic ecosystems, redefining productivity, pricing models, and user experiences.

‍

Deploying Agentic AI: Lessons from Early Wins and Setbacks

A year into the agentic AI revolution, one truth stands out: success doesn’t come easy. Some companies are seeing productivity gains, but many are struggling to realize value or even retrenching when deployments fail. McKinsey’s analysis of over 50 agentic AI builds they led, along with analysis of dozens of others in the marketplace, highlights key lessons on how to get agentic AI right.

It's about the workflow, not just the agent

Achieving value requires rethinking workflows—people, processes, and technology together—not just building impressive-looking agents. For example, legal-services providers and insurers that built feedback loops into workflows enabled agents to learn continuously and improve results.

Agents aren’t always the solution

Too often, leaders overapply agents when simpler tools would suffice.

‍

• When tasks are highly standardized → rules-based automation or predictive models may work better.
• When tasks are variable and complex → agents can add real value.
• Best practice: treat agents like team members, assigning them only where they fit.
  ‍

Reuse Beats Reinvention

‍Instead of spinning up a new agent for every task, leading firms invest in reusable components—validated prompts, reusable code, and common orchestration frameworks. This reduces redundancy, accelerates deployment, and eliminates waste.

Humans Remain Central

‍Even as use of AI agents expand, humans remain vital for oversight, judgment, and compliance. The most effective deployments:

• Are designed for human–agent collaboration.
• Use intuitive interfaces to build trust and speed adoption.
  Without this, even advanced systems risk user rejection.

The Road Ahead

‍Agentic AI is advancing quickly, but one lesson is clear: success depends less on the technology itself and more on how organizations redesign workflows, choose the right tools, build user trust, and thoughtfully integrate humans and agents.

‍

Innovation Rewired: Where Human Imagination Meets AI

AI can’t dream, but it can supercharge the path from idea to market. Bain’s Innovation Report 2025 shows how the smartest companies are blending machine efficiency with human imagination.

​Innovation is still broken

Companies run hackathons, fill whiteboards, and launch idea challenges—yet most projects stall. Only 5–25% of new products actually succeed. Systems are still noisy, slow, and wasteful.

​Where AI helps most  

AI isn’t replacing creativity—it’s speeding up everything around it. Leading firms are using AI to scan massive datasets for trends, build and test prototypes virtually, predict which ideas are most likely to succeed, and automate the tedious parts of R&D so humans can focus on strategy.

Already, 88% of top innovators say AI has improved their innovation success rates, with many cutting design-to-launch times by more than 20%.

​Where AI falls short

Despite these gains, AI still struggles in the areas that matter most for real breakthroughs. It finds it hard to generate truly radical, out-of-the-box ideas and tends to avoid risky bets, sticking too closely to historical data. It also often misses the nuances of human empathy and cannot replicate the messy, serendipitous collaboration that sparks creativity in human teams. In short, AI can sharpen the process, but it cannot replace the spark of human originality.

​The human-AI balance  

The lesson from Bain’s case studies is clear: growth comes when AI and humans co-pilot. A US life insurer, for example, used AI analysis and synthetic customers to redesign its products strategy, unlocking a potential $1B business. Similarly, a telecom giant paired AI with traditional research to crack new customer segments without cannibalizing its premium brand. These examples show that the real magic happens when companies let machines accelerate the process while humans shape the vision.

‍The next chapter  

AI is here to accelerate, not imagine. True breakthroughs will still come from humans—dreaming big, taking risks, and steering strategy. The future belongs to companies that know when to let AI run, and when to let human creativity lead.

‍

That’s it for the September’s edition of This Month In AI. We hope you enjoyed the read.

​Sign up for our monthly newsletter to get this curated list of AI articles and more AI insights from RapidCanvas delivered straight to your inbox.