RightTouch, Inc.

・Incident Lake

RightTouch, Inc. is an up-and-coming startup that provides AI platforms to enterprise companies with the goal of transforming the contact center industry.The company faced challenges such as strained resources for incident response due to rapid growth and the time-consuming nature of investigating root causes caused by increasingly complex systems. Consequently, the company decided to implement SIGQ’s AI agent platform, “Incident Lake.” By consolidating information across organizational silos and enabling automated reporting, the company aims to eliminate the reliance on individual expertise in incident response and build a system that generates greater value with limited resources.

■Challenges Before Implementation

• When an outage occurred, customer engineers spent a significant amount of time gathering technical information and reporting it to customers.

• The quality of incident reporting depended on individual skills and circumstances, resulting in inconsistencies in the level of detail and wording of the reports.

• With the infrastructure and repositories of our company and our group companies intermingled, identifying the cause of issues had become increasingly complex

■Expected Benefits of Implementation

• Incident Lake’s information aggregation and automated reporting capabilities reduce the time and effort required to interpret scattered technical information and draft reports, enabling organizations to focus on higher-value-added tasks.

• By standardizing processes using AI, we can unify reporting formats and terminology, enabling the rapid submission of high-quality incident reports that are free from individual bias.

• Incident Lake’s cross-organizational knowledge base eliminates silos, structures complex system information, and facilitates root cause analysis and rapid decision-making.
  

Behind the rapid growth of AI-powered customer support, the workload associated with handling service outages is skyrocketing

RightTouch Inc. is a startup established in December 2021 as a spin-off from PLAID. Centered around its web support platform "QANT Web," the company offers a range of products designed to transform contact centers through the full utilization of AI, including "QANT VoC," which improves customer service based on VoC (Voice of Customer), and the AI voice operator "QANT Speak."

The company's system infrastructure employs a multi-cloud configuration using GCP and AWS. Due to its origins as a spin-off from PLAID, the infrastructure is broadly divided into two layers.

The first is the KARTE platform provided by PLAID. We use this KARTE platform to track end-user web behavior on our web support platform.

The second is an in-house platform developed exclusively by RightTouch. This platform stores customer support data related to inquiries—such as Voice of the Customer (VoC) feedback from phone calls, the context of inquiries, and details of interactions—which cannot be measured through web behavior. Another key feature of the company’s technical architecture is that applications run on both the KARTE platform and the company’s proprietary infrastructure.

Furthermore, the company continues to pursue cutting-edge initiatives in AI, including the integration of large language models (LLMs) into products centered around Gemini, the use of AI for data analysis such as VoC, and the AI-native development of new products.

As the company experienced rapid business growth, it faced two major challenges: a shortage of resources for incident response and siloed operations resulting from system complexity.

The first challenge is the strain on resources for incident response. When an incident occurs, customer engineers take the lead, collaborating with product and business teams to resolve the issue and handle customer inquiries. However, with a limited number of customer engineers, handling sudden incidents while simultaneously providing support for other products placed a significant burden on them.A particularly time-consuming task was deciphering technical information flowing through various channels, filtering out the necessary facts, and then reporting them to customers in plain language. As a result, the strain on resources created a vicious cycle in which proactive customer support and the development of delivery systems for business expansion were inevitably put on the back burner.

The second issue was siloing caused by system complexity. As mentioned earlier, the company’s systems span multiple platforms, and source code repositories and Slack channels are distributed across various organizations. As a result, isolating the root cause of issues was time-consuming, and in some cases, it took a long time to resolve problems unless the team included members with experience at both PLAID and the company.This complexity, both in the systems and the organization, hindered information sharing and led to a reliance on specific individuals for incident response.

In light of these challenges, Mr. Haruki Kawashita, a customer engineer at the company, had the following to say about the ideal operational model:

"Customer engineers are rare professionals who bridge the gap between technology and business. To maximize limited resources, we must drive operational automation so that we can focus on areas requiring human intervention and concentrate on optimizing AI products and delivering value."

Gather scattered information and report it in terms that customers can understand

To address these challenges, RightTouch decided to implement SIGQ’s AI agent platform, “Incident Lake.”

Mr. Kawashita has the following four expectations for Incident Lake:

• Compiling technical information scattered across various locations and channels and accurately identifying the facts

• Reporting based on those facts using language that customers can understand

• Preventing situations where the cause cannot be identified without relying on a specific individual

• Ensure that post-incident reviews are conducted to prevent recurrence

“Currently, we spend a significant amount of time compiling information and reporting to customers. From the perspective of reliability, it is critically important to determine how much we can reduce the time required to post updates on the status page and send initial notifications to customers, as well as how we can ensure the accuracy of that information,” said Mr. Kawashita.

How will Incident Lake meet these expectations?

Incident Lake features a system that groups data sources from multiple organizations, sets permissions, and automatically collects and shares only the information relevant to a specific incident. This approach enables both the aggregation of information across organizational silos and permission management that ensures only necessary information reaches the relevant parties. As a result, even when Slack workspaces or GitHub repositories are divided across organizations, Incident Lake can powerfully aggregate information and eliminate silos.

Another key feature is the structuring and accumulation of knowledge through a knowledge base.It features a mechanism where AI extracts tacit knowledge—which is often tied to specific individuals—and converts it into explicit knowledge, eliminating the need for humans to manually write documentation. For example, you can save knowledge directly via Claude Code MCP or set up a system to periodically update saved knowledge using GitHub Actions. This ability to automate knowledge management without human intervention is a major advantage of Incident Lake.

The company plans to gradually transform its incident response processes through the implementation of Incident Lake. Mr. Kawashita explains that the first step will be to input the necessary knowledge and historical incident data, and then utilize Incident Lake in actual incident response while continuously fine-tuning the system. Additionally, the company intends to work on standardizing reporting quality through AI and accelerating decision-making during the initial response phase.

“Our ideal scenario is to utilize Incident Lake across the entire incident response operation, creating an environment where even new team members can handle incidents smoothly simply by interacting with Incident Lake’s AI—without having to refer to individual documents from various sources,” said Mr. Kawashita.

Overcoming the Temptation to "Do It Ourselves"

Before implementing Incident Lake, RightTouch had also considered developing a solution in-house. This was because the company already had several AI engineers on staff and was actively working to optimize operations using AI—for example, by setting up a system that allowed them to easily retrieve source code from GitHub and ask questions via Slack.

However, Mr. Kawashita explains the reason for not opting for in-house development as follows.

“Even if we were able to build our own tools through AI-driven development, they wouldn’t be fit for practical use without ongoing fine-tuning. Furthermore, practical experience in troubleshooting is essential for developing the features that operations teams actually need. As a result, this could end up taking up a significant amount of our engineers’ time,” said Mr. Kawashita.

Mr. Kawashita reflects that the deciding factor in adopting Incident Lake was the conclusion that it offered better value for money to implement a product refined by a specialized company with expertise in incident response, rather than spending time and manpower developing a tool that would have to prove its practicality.

Unlimited scalability to support lean, high-performing teams

Regarding the future outlook for the company’s customer engineers, Mr. Kawashita states that he wants to focus on providing advanced support to help deliver the value of AI products to customers.Specifically, this involves having specialists with a deep understanding of customers’ day-to-day operations perform optimal tuning for each customer, thereby helping them unlock the full value of AI products. Mr. Kawashita’s vision is to contribute not only to business growth but also to the refinement of the products themselves. To achieve this, he expresses his ambition to focus on recruiting highly skilled talent who are well-versed in both business and technology, despite the scarcity of such professionals.

Finally, Mr. Kawashita offers some advice to companies facing similar challenges.

“My honest opinion is that it’s better to simply adopt SaaS without hesitation rather than going through the trial-and-error process of managing it ourselves. I believe the incident response process shares many common elements across all companies. It’s more efficient to align our company’s methods with best practices that deliver consistent performance, and we can expect even better results,” said Mr. Kawashita.

An increase in headcount does not necessarily correlate with business growth. This is particularly true for startups experiencing exponential growth, where SaaS—which offers unlimited scalability—can be a powerful tool for addressing staffing shortages. SIGQ will continue to support the company’s rapidly growing product through Incident Lake.