How DMS One put Claude to work across the whole company, from leadership to the codebase

With Claude, DMS One:

• Cut the turnaround time of critical pre-sales processes by 40 to 60 percent
• Reduced administration time by 13 percent by feeding meeting intelligence directly into CRM activity logging and customer analysis
• Operates more than 10 reusable agentic workflows in production across pre-sales, development, support, and project management
• Prepares a classification and priority recommendation for incoming support tickets, with a written justification, so the support team decides with full context
• Gives developers a head start on new development and bug tickets, with Claude Code preparing an analysis and a suggested fix that engineers review, adapt, or reject
• Brought more than 20 people to confident daily use through a structured training program

The challenge

DMS One builds and operates document and workflow software for more than 3,500 companies and public institutions across Hungary, with a lean team covering delivery, support, engineering, and leadership. That model concentrates a great deal of recurring, knowledge-heavy work on a small number of people. Every support ticket had to be read, categorized, prioritized, and estimated by hand. Consultants and support staff had no fast way to trace system behaviour across a large codebase. Leaders assembled oversight reports manually, and project managers rebuilt status summaries and meeting preparation every cycle. Where AI was used at all, it lived in isolated pockets.

The solution

As DMS One's Delivery Partner of Record in the Claude Partner Network, BPDATA set out to make Claude a single platform the whole company could rely on, rather than a tool for one team.

At the centre is an MCP-based integration architecture connecting Claude to DMS One's core systems, including Jira, Bitbucket, Salesforce, and Google Workspace. Incoming support tickets receive a classification and priority recommendation with a written justification, and a scheduled daily report orders the open queue by urgency; the support team reviews every recommendation and makes the final call. Consultants and support staff ask about system behaviour and errors in plain language and get answers grounded in the live code. For new development and bug tickets, Claude Code prepares an analysis and a suggested fix as a starting point, which engineers review, adapt, or reject. Throughout, the AI acts as decision support: people stay in control of every classification, priority, and change that ships.

The integrations were built with data protection as a design requirement. Access to each connected system is scoped to what the given workflow needs, personal data is handled in line with GDPR, and every AI output that affects a customer or the codebase passes through human review, consistent with the human oversight principles of the EU AI Act.

The pre-sales specification and prototyping process was rebuilt as an agentic workflow connected across the wider toolchain, cutting the turnaround time of critical pre-sales processes by 40 to 60 percent. Meeting intelligence flows directly into CRM activity logging and customer analysis, reducing administration time by 13 percent.

For the leadership team, scheduled Cowork jobs handle recurring oversight and activity monitoring, and project managers receive current-status summaries and meeting preparation on a set schedule. The rollout went function by function, from the consulting and project team to support, product, and testing, underpinned by a library of internal Claude skills and a structured training program that brought more than 20 people to confident daily use.

The outcome

Claude is now in production at DMS One, in daily use by 24 people spanning leadership, consulting, project management, support, product, and testing rather than a single team. Critical pre-sales processes run 40 to 60 percent faster, administration time is down 13 percent, and more than 10 reusable agentic workflows carry recurring work across the company. Support tickets arrive with a documented recommendation, development tickets arrive with a prepared analysis and suggested starting point, and leaders and project managers start from prepared reports instead of a blank page, with people making the decisions at every step.

Looking ahead

For DMS One, the internal rollout was the first step, not the destination. Existing assistants are being consolidated onto the platform, new internal agents are being built there from the start, and the skill library grows month by month. Building on the results of the internal deployment, the company is also evaluating AI-based features in the DMS One product itself.

Delivered by BPDATA Kft., Delivery Partner of Record, Claude Partner Network.