GBG launches Reach to verify AI agents' address data

GBG has launched Reach, a data quality agent for AI agents. The product integrates with workflows including IBM watsonx Orchestrate.

Reach is the first offering in GBG for Agents, a new portfolio designed to give AI systems policy-based decision support when handling address information. It verifies address data in real time and returns a confidence score from 0.0 to 1.0, allowing agents to decide whether to proceed.

The launch comes as businesses try to move AI agents from pilot projects into routine operations. GBG cited Gartner research forecasting that 40% of enterprise applications will feature task-specific AI agents by the end of 2026, up from less than 5% in 2025.

Data quality remains a barrier. GBG also pointed to McKinsey research showing that 80% of organisations cite data limitations as an obstacle to scaling agentic AI.

The issue is particularly acute in processes where poor address data can create knock-on problems. Unverified address information can introduce risks in fulfilment, customer onboarding and identity-related checks when AI agents are allowed to act without clear validation.

Decision layer

Rather than requiring customers to build their own verification logic into agent workflows, Reach is intended to act as a decisioning layer between the workflow and the underlying address data. In practice, an AI agent can submit an address, receive a verification outcome and confidence score, then continue a task based on those results.

The approach is designed to make agent behaviour more consistent and easier to govern. Existing Loqate customers can also activate Reach using their current API key under a bring-your-own-key model, avoiding a separate purchasing process.

Kartik Venkatesh outlined the company's view of how these systems are developing across business processes.

"AI agents are going to orchestrate more and more of the identity and location journey - onboarding, checkout, fulfilment, verification, screening - increasingly driven by multi‐agent systems. But somewhere in that chain, someone still needs to make a trusted decision with real data underneath. That's where GBG sits, not replacing the agentic journey, powering it. With GBG for Agents, starting with Reach, we're making our market-leading verification and decisioning capabilities composable, agent-native, and ready to plug into workflows starting today," said Venkatesh, Head of Global Innovation, GBG.

Address focus

The product centres on address verification, an area where GBG already has an established presence through Loqate. By turning that existing service into a format suitable for AI agents, GBG is seeking to position itself within a growing software market around autonomous and semi-autonomous business systems.

For organisations deploying AI agents, address verification may appear narrow compared with broader generative AI functions, but it sits close to revenue and risk. Errors in delivery details can affect logistics and customer service, while incorrect address data in onboarding or identity workflows can undermine compliance checks and fraud controls.

IBM watsonx Orchestrate is one of the workflow environments named by GBG, suggesting Reach is intended to sit inside larger process automation stacks rather than operate as a standalone application. That reflects a wider shift in enterprise software, where suppliers are packaging narrowly defined functions for use by AI systems that trigger actions across multiple tools.

Broader push

Reach also marks a broader product move for GBG as it adapts its identity and location tools for agent-led environments. The company plans to extend the GBG for Agents portfolio to other parts of its product range, including GBG Go and GBG Trust.

GBG operates in identity and location technology and says it serves more than 20,000 customers globally. It is listed in London and is a constituent of the FTSE 250.

The commercial case for products like Reach will depend on whether businesses adopt AI agents widely enough to justify dedicated controls around the data those systems use. But the launch underlines a growing focus in enterprise AI on the quality of operational data, not just the sophistication of the models making decisions.