ShipStation Global reports Deckhand handled 97.5% of Slack IT messages
ShipStation Global reports Deckhand handled 97.5% of Slack IT messages
AI Agents & Automation

ShipStation Global reports Deckhand handled 97.5% of Slack IT messages

AWS Machine LearningOct 13, 2026

ShipStation Global says its Slack agent Deckhand answered 97.5% of messages and earned 96% positive feedback during its first three months.

ShipStation Global says its AI help-desk agent, Deckhand, handled 97.5% of messages in the company’s IT support Slack channel during its first three months. In an AWS-published account dated October 13, 2026, the company reported a 96% positive employee feedback rating in June, up from 82% in April.

Launched in April, Deckhand had handled 1,935 messages and started 1,221 agent conversations by the end of its first three months. Of those conversations, 409 resulted in Jira tickets; the remaining 812 were resolved in chat, according to the account. ShipStation estimates that avoiding those tickets saved about 406 hours over the period, based on roughly 30 minutes to file, work and close each one. It says that a third of questions arrived after hours, and the agent answered 94% of those.

Deckhand runs in Slack using Amazon Bedrock AgentCore and a Strands agent with Claude on Amazon Bedrock. It draws on a knowledge base that includes Confluence articles, company intranet materials and resolved support threads. It can also inspect screenshots of error messages and use connected tools to classify requests and create or update Jira Service Management tickets. If it lacks an answer, it can ask a follow-up or offer to prepare a ticket.

The system uses rules to avoid interrupting social conversations or responding after a human colleague has taken over. Ticket creation requires an employee’s confirmation, and AWS says credentials are kept in Secrets Manager with service-specific access controls. For its knowledge base, ShipStation processes past Slack threads to distinguish documented resolutions from unresolved conversations and other channel activity.

The account describes a company-specific deployment rather than an independent evaluation. ShipStation reported more than 1,500 support threads in the 12 months before launch, with questions spread across several internal sources and arriving across multiple time zones. Its reported results reflect the first three months of Deckhand’s operation.

Industry Impact

AI takes on routine IT support

Enterprise IT support and business operations

Overview

ShipStation Global’s first three months with Deckhand show how an AI agent can handle a large share of requests in an internal Slack help desk. The agent handled 97.5% of messages, resolved 812 conversations in chat and routed 409 to Jira tickets. Employee feedback rose to 96% positive in June, from 82% in April. For business leaders and AI implementation teams, the results suggest a way to extend support across time zones and after hours, while security teams should note the company’s confirmation and access-control safeguards.

Key Impact Points

812 conversations were resolved in chat, and ShipStation estimated avoiding 406 hours of ticket work.

The agent answered 94% of after-hours questions, extending support beyond staffed hours.

Ticket creation required employee confirmation, with credentials stored in Secrets Manager and service-specific access controls.

Opportunities

Automate routine help-desk conversations

Enterprise IT support and service management

Organizations handling recurring employee IT questions in Slack or similar channels could use an agent connected to approved knowledge sources and ticketing tools. Deckhand’s workflow provides a concrete pattern: answer from documentation and resolved support threads, ask follow-up questions when needed, and prepare or update tickets when chat resolution is insufficient. A human confirmation step for ticket creation and rules against interrupting active colleague responses can help retain oversight. The reported 406 hours saved is an estimate based on ShipStation’s assumptions, not a guaranteed result for other organizations. Any deployment should validate answer quality, escalation behavior, security controls and employee feedback in a limited pilot.

  • Slack-based IT help-desk agent

    Problem: Employee IT questions arrive across internal sources and time zones, creating support threads and manual ticket work.

    Solution: Connect a Slack agent to approved documentation, resolved support conversations and ticketing tools; allow it to answer, ask follow-up questions and prepare tickets for employee confirmation.

    Target Users: Internal IT support teams and employees seeking help.

    Business Value: Potentially resolve routine questions in chat and reduce manual ticket workload; ShipStation estimated 406 hours saved in its first three months.

    Revenue Potential: No direct revenue model is reported. The demonstrated value is internal time savings, and results may vary by organization.

Reference link

https://aws.amazon.com/blogs/machine-learning/how-shipstation-global-built-an-ai-it-help-desk-on-amazon-bedrock-agentcore

Image Source Link

https://d1.awsstatic.com/onedam/marketing-channels/website/aws/en_US/homepage/global-nav/reinvent-register.3822079bac49f139ae46016a9a99c3eed71fad4a.png

Community Discussion (0)

Loading discussions...