Why Custom AI Agents Beat Enterprise SaaS Software

I have noticed a growing gap between what leaders expect from AI and what it actually delivers. While many assume technology makes everything immediate, a 2025 MIT study titled "The GenAI Divide" reported that 95% of corporate AI pilots are failing to deliver a measurable impact on profit and loss. This creates a "trough of disillusionment" where the expectation for success is much higher than reality. I see this putting immense pressure on research teams. Executives are often pushed with the idea that AI should make tasks automatically easier, leading to a conversation about what is "good enough." Some companies are now debating if 70-80% accuracy is acceptable, but for a leader, making decisions based on incomplete certainty is a major risk.

Recently, I spoke with an organization regarding their AI adoption. Like many, they provided access to tools like Copilot or Gemini and essentially told their team to "figure it out." While corporations will eventually find broad solutions to sell as SaaS products, there are many unique, single tasks performed by employees that do not justify a massive enterprise expense. These specific, custom needs often remain unmet by one-size-fits-all software.

The biggest challenge I see is data quality. Currently, about 20% of collected data is being thrown out and replaced because it is unreliable. While AI is part of the solution, it is also part of the problem. Experts agree that human review is the only way to ensure quality, even if it makes the process take longer. At Team Percepto, I believe that before spending on new studies, you must tap the internal data you already have—from sales logs to customer reviews. As I noted in my work on Human-First Full Spectrum Data Insights, the skill lies in pulling signals from multiple sources, using custom tools to process them, while human logic determines the meaning.

I believe we are at a point where the need for "humanness" is stronger than ever. History shows that when research moved to the internet years ago, it actually increased the desire for in-person qualitative work. True technology adoption is not just about the tool; it is about the discipline of the person using it. In my experience with Personal AI Development, I have found that you must move beyond using AI as a search engine. To succeed, you have to be the creator of your own agents, fine-tuning them to your specific processes rather than letting them replace your judgment. Even when an internal prompt runs autonomously, it still requires human-centric prompting to remain accurate.

My goal is to help you take confident action by balancing these tools with human-led analysis. As discussed in our AI Adoption Strategies, success requires a "creator mindset" where humans provide direction and technology acts as the enhancer. Team Percepto can help build solid, custom agents tailored to your specific needs without the burden of enterprise SaaS licensing. By reviewing every research process, we find ways to save time without losing the human touch. Consumer research should clarify your questions and provide powerful insights that help you win.

Custom AI agents

By Juan Jose (JJ) Ayala

Team Percepto, A Consumer Research Insights & AI Adoption Company, February 2026