Answers from the documentation; checking which model is on duty
Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
AT&T adopts agentic AI to transform business workflows - OpenSmartRoute
AT&T adopts agentic AI to transform business workflows
AT&T uses over 500 autonomous workflows powered by agentic AI, aiming to triple that number next year. The approach focuses on redesigning entire processes.
Key points
Nearly 500 agentic workflows are in production at AT&T.
The company processes about 45 billion AI tokens daily.
AT&T’s platform handles over 1 million transactions monthly.
Success metrics include fivefold ROI within a year.
Why it matters: Automating entire processes with agentic AI can reduce costs, speed up services, and improve customer satisfaction.
By OpenSmartRoute editorial · written through the router by llm-small
From SiliconANGLE AI - “AT&T uses agentic AI to reinvent business processes from the ground up”
Introduction - AT&T’s shift to agentic AI for business process redesign
AT&T is moving beyond traditional AI tools to use autonomous agents for redesigning its business processes. The company now focuses on creating workflows that operate independently. This shift aims to make operations more efficient and flexible. It marks a change from automating individual tasks to rethinking entire processes.
This approach involves building systems that can manage complex tasks without constant human input. AT&T’s goal is to improve speed, reduce costs, and enhance service quality. The company’s strategy emphasizes redesigning how work gets done, not just automating parts of it. This change reflects a broader trend toward agentic AI in large organizations.
Current scale and scope of AT&T’s agentic workflows
AT&T already has nearly 500 agentic workflows in production. These workflows handle over 1 million transactions each month. They process about 45 billion AI tokens daily across the company. A token is a piece of data used by AI models to understand and generate language.
The workflows cover a wide range of tasks. They include customer provisioning, sales prospecting, and interactions with legacy systems. Many of these workflows now operate autonomously, without direct human control. The company plans to triple the number of agentic workflows next year.
This scale shows how deeply AT&T has integrated autonomous agents into its operations. The company sees these workflows as essential to its digital transformation. They aim to replace manual, time-consuming tasks with faster, automated processes.
Details of the agentic AI platform and architecture
AT&T’s platform for agentic AI is built around a small set of central agents. These agents execute business workflows using shared logic stored in memory. Instead of many specialized agents, the company maintains a few that handle multiple tasks.
The architecture includes retrieval agents and planning agents. Retrieval agents find relevant knowledge needed for a task. Planning agents decide what actions, tools, and data are necessary to complete the task. This setup allows the system to be flexible and adaptable.
LangChain, an AI agent tooling startup, secured $300 million in funding, boosting its valuation to $3 billion. It plans to grow its team and expand its open-source and paid products.
Security and governance are key parts of the platform. Workflows that affect customers must go through reviews for security, privacy, and compliance. Employees can create personal productivity agents, but workflows with customer impact are controlled. Human oversight is required until agents reach high accuracy levels, typically in the mid- to high 90% range.
Examples of process automation and improvements
One early example involved automating interactions with AT&T’s mainframe computers. Previously, employees had to manually retrieve tickets, update records, and return them to the workflow. This manual process caused delays and inefficiencies.
Agentic automation eliminated parts of this process. It reduced delays and sped up service delivery. Employees no longer needed to handle routine data updates, freeing them for more complex tasks. The automation also improved accuracy and consistency.
Another example is the halving of the time needed to provision dedicated internet connections. When fiber infrastructure was already in place, the process used to take longer. Now, with agentic workflows, the setup is faster, reducing customer wait times. These improvements show how autonomous agents can streamline operations.
Governance, security, and oversight of workflows
AT&T emphasizes strict governance for its agentic workflows. Workflows that impact customers must pass through a review process. This process checks for security, privacy, and compliance issues before deployment.
Employees can create personal productivity agents freely. However, workflows that could affect customers or the business must be approved first. This approach balances innovation with risk management. Human oversight remains part of every production workflow until the system proves reliable.
Continuous monitoring is used to evaluate agent performance. The system tracks accuracy and detects any drift from expected behavior. Only when agents reach a high level of accuracy can they operate without human oversight. This process ensures safety and reliability in the workflows.
Metrics of success and business impact
AT&T measures success through several metrics. These include labor savings, revenue growth, operating costs, and customer satisfaction. The company reports that its AI investments are delivering a fivefold return within a year.
Reducing manual work lowers labor costs and improves efficiency. Faster processes lead to quicker service delivery and higher customer satisfaction. Revenue increases from better sales prospecting and faster market entry. These metrics help AT&T justify ongoing investments in agentic AI.
The focus on measurable outcomes ensures that the company’s AI efforts align with business goals. It also provides a clear way to track progress and identify areas for improvement.
Challenges in scaling and data infrastructure
Scaling agentic workflows across a large organization like AT&T presents challenges. The company has gone through many mergers and acquisitions, which created complex data infrastructure. Consolidating fragmented data sources is necessary for reliable AI performance.
Providing consistent connectivity across data centers, cloud platforms, and mobile networks is also critical. These infrastructure issues can slow down deployment and affect the quality of AI services. Addressing these challenges requires careful planning and investment.
Data security and privacy are additional concerns. Ensuring that sensitive customer data remains protected while enabling AI workflows is vital. AT&T’s governance framework helps manage these risks, but scaling remains a complex task.
Future plans for process automation and organizational distribution
AT&T plans to expand its use of agentic AI significantly. The company aims to triple the number of workflows next year. It also intends to use agents more broadly to launch new services, enter markets, and support sales channels.
The company is shifting from a centralized team to a more distributed model. Responsibility for AI-driven processes will spread throughout the organization. This approach encourages employees at all levels to develop and deploy workflows.
The goal is to embed agentic AI into everyday operations. By distributing responsibility, AT&T hopes to increase agility and innovation. This strategy aims to make AI a core part of the company’s organizational culture.
How OpenSmartRoute helps with agent management and routing
OpenSmartRoute is an open-source router designed for AI requests. It directs each request to the best-fit model, agent, tool, or skill from a defined catalog. The router scores candidates based on quality, cost, speed, and safety. The team sets the importance of each factor for every request.
OpenSmartRoute ensures requests follow rules, such as keeping personal data on on-premises models or respecting regional limits. It learns from outcomes, giving more traffic to models that perform well. If a new model is added, it competes with existing options on each request.
The platform also provides a savings ledger. It shows how much each request costs compared to the most expensive option. The 'osr eval' tool measures routing accuracy on the team’s prompts. It can flag poor performance and prevent deployment of unreliable models.
For teams adopting agentic AI like AT&T, OpenSmartRoute offers a way to manage and optimize request routing. It helps ensure that workflows operate efficiently, securely, and cost-effectively. This supports scaling and governance as autonomous processes grow.
How it compares - what existed before, what this changes and what stays the same
Before AT&T’s move to agentic AI, the company relied on traditional automation and manual processes. These methods involved automating individual tasks or using scripted workflows. They often required human intervention and were limited in scope.
Automation previously focused on replacing repetitive tasks. It did not aim to redesign entire processes. Many workflows still depended on legacy systems and manual handoffs. This limited efficiency gains and made scaling difficult.
The new approach shifts from automating isolated tasks to redesigning whole processes around autonomous agents. AT&T now uses a small set of centrally managed agents that execute complex workflows. These agents use stored business logic and can identify the right actions and data needed.
This change allows for more flexible, scalable, and intelligent workflows. It reduces delays, eliminates unnecessary steps, and minimizes dependencies on legacy applications. The company can now handle more transactions and processes with fewer human workers.
What stays the same is the core goal of automation: improving efficiency, reducing costs, and enhancing customer experience. The company still values governance, security, and oversight. Human oversight remains part of the process until agents reach high accuracy levels.
This shift also maintains the importance of data infrastructure. AT&T continues to consolidate fragmented data and ensure connectivity across systems. These foundational elements support the new agentic workflows.
In summary, AT&T’s move from task automation to process redesign with agentic AI changes how workflows are built and managed. It replaces many manual steps with autonomous agents, while keeping the focus on efficiency, security, and governance.
Questions this leaves open - what the source does not say and how a reader can check it
The source does not specify the exact size of the agentic AI platform. It mentions nearly 500 workflows in production and plans to triple that number, but does not detail the total number of agents or models involved.
It is unclear how the company measures the success of these workflows beyond broad metrics like labor savings and customer satisfaction. Details about specific evaluation methods or benchmarks are not provided.
The source does not describe the technical architecture in detail. For example, it does not specify the types of models used, how they are integrated, or what infrastructure supports the platform. Nor does it explain how the platform handles failures or errors.
Questions also remain about the governance framework. The source mentions security, privacy, and compliance reviews, but does not clarify how these are implemented at scale. It is not clear how the company manages risks associated with autonomous agents operating in live environments.
The source does not specify the costs involved in building and maintaining the agentic AI platform. It also does not compare these costs to previous automation investments or traditional IT infrastructure.
To verify or learn more, a reader can explore AT&T’s public statements or technical reports. They can also review case studies or white papers on agentic AI in large enterprises. Attending industry conferences or webinars featuring AT&T’s digital transformation efforts may provide additional insights.
Another approach is to examine similar implementations in other companies. Comparing approaches can reveal common challenges and best practices. Engaging with AI and automation communities can also help clarify how such systems are built and governed.
Finally, testing or piloting small-scale agentic workflows within one’s own organization can provide practical experience. This helps understand the technical requirements, governance needs, and potential benefits or pitfalls.
In conclusion, while the source provides a broad overview of AT&T’s strategy, many technical and operational details remain open. Further research and direct engagement are necessary to fully understand the scope and implications of agentic AI at scale.
Mercor announced a $750 million funding round, valuing the company at $20 billion. The money will boost infrastructure and research for AI model evaluation and training.
