July 2026 Summaries

This text discusses the common questions users ask AI assistants integrated into software products, highlighting recurring themes across various applications. Frequent inquiries include financial matters such as billing and refunds, user orientation within the software, permissions and roles, understanding metrics on dashboards, and issues related to setup and integration. Users often seek more control over existing features rather than entirely new capabilities, and offboarding processes receive little attention despite generating numerous questions. The text emphasizes that these issues are not unique to poorly designed software but are inherent challenges as products evolve and expand their functionalities. It suggests that in-product AI assistance offers immediate, context-specific help, providing a more effective solution than traditional help documentation and support queues, which often lag behind in addressing user confusion.

In 2026, the customer support AI landscape is characterized by a variety of tools designed to address specific needs based on the channel they operate within, such as in-product assistance or support inboxes. Among these, Frigade stands out as an AI that runs within products to help users navigate complex features, effectively reducing the need for support tickets by facilitating direct problem-solving. In contrast, tools like Intercom Fin and Zendesk AI Agents enhance support chat and ticketing platforms, respectively, offering solutions for teams already using these systems by leveraging existing help documentation. Ada and Forethought cater to enterprise customer experience teams and Salesforce-heavy environments, respectively, with a focus on vendor maturity and intent classification. For high-volume consumer or SMB support, autonomous agents like Decagon and Sierra are recommended, while Pylon offers a tailored solution for B2B SaaS companies conducting support through shared Slack channels. The decision-making process for selecting among these tools involves analyzing the source of support volume, customer base characteristics, and whether the goal is to answer existing tickets or prevent them from being created.

Frigade has introduced a major update to its in-product assistant, known as Skills, which allows the assistant to perform actions within the product, marking a significant advancement from merely providing instructions. This update enables the assistant to autonomously learn and execute tasks such as lookup, create, edit, and delete operations by observing user interactions, thus minimizing the need for manual integration efforts. Skills are only activated after user approval to ensure security and maintain control, with actions being confined to what users can already access within the product. This development is in line with the evolving user expectations of having agents perform tasks directly, rather than just explaining processes. Skills aim to enhance user experience by allowing the assistant to handle repetitive tasks, thus setting a new standard for automated assistance in dynamic product environments.

In a field report detailing a year of usage data from Frigade Assistant, an in-product AI for SaaS, key insights reveal its effectiveness and unique capabilities compared to traditional chatbots. Unlike chatbots that rely on static knowledge bases, this AI operates within the live product, offering guided walkthroughs that engage users deeply, with 62% of returning users asking three or more questions, and about 1 in 10 asking eleven or more. Approximately 35% of the assistant's responses are guided walkthroughs, which are followed by users 70% of the time, demonstrating a significant engagement rate as users are walked through workflows on their current screens. This proactive, integrated approach reduces reliance on external help centers and positions in-product AI as a more intuitive and efficient solution for user support and task completion, with its success measurable by the rate of users returning to ask subsequent questions rather than just the volume of queries.