AI Agents for Information Services: What, Why and How to Succeed

Agentic AI is the numero uno technology trend for 2025, according to Gartner. The public interest in AI agents has also gone up significantly. Mention of ‘agentic AI’ in earnings calls grew 50x since Q1, 2022; GitHub repositories increased more than 150x, finds Bain & Company.

But what is agentic AI? How does it work? Is it worth the hype? That’s what we seek to find out in this blog post. However, for better focus, we will keep this discussion on AI agents for information services.

What is agentic AI?

Agentic AI refers to autonomous robots that can complete tasks independently, without human intervention, adapting to changing contexts. This means that an AI agent can:

• Understand your goal and work toward it (instead of just executing pre-defined tasks)

• Perform complex tasks that need multiple steps or coordinating with multiple agents

• Integrate with various enterprise tools for data or actions

• Handle changes in data or other input and adapt accordingly

When we talk about gen AI agents, we mean software entities that can orchestrate complex workflows, coordinate activities among multiple agents, apply logic, and evaluate answers.

- Lari Hämäläinen, Senior Partner, McKinsey Digital

Why are AI agents better?

Robotic Process Automation (RPA) was traditionally rule-based. Even intelligent automation demanded a clear statement of the process, including all exceptions. Chatbot-based tools like ChatGPT need prompting for the AI to work. 

Agentic AI takes the collective capabilities of generative AI and automation one step further than traditional AI systems.

Specialization: One of the problems leaders face with generalized AI solutions is that their knowledge goes wide but not deep. Agentic AI solves that problem. It enables you to create AI agents for information services that can each specialize in their area of expertise and then work together to complete a task.

Adaptability: Modern AI agents can adapt to changes in the environment. They are great for workflows that are less predictable or have a large number of potential outcomes. Agentic AI can apply logical reasoning and make complex decisions. Most importantly, AI agents continuously learn from experience.

Integrations: Enterprise Resource Planning (ERP), customer relationship management (CRM), payment processing, finance, workforce planning, inventory management — modern AI agents for information services can integrate with all these and more to complete tasks.

Ease of use: Unlike the RPA of the previous generation, agentic AI leverages natural language processing to understand user input in plain English. Business users can state what they want in plain English, which the AI will convert into workflows. 

The benefits of efficiency, productivity, speed, scale and saving operational costs go without saying!

Why is everyone talking about AI agents for information services?

Before we get into answering that question, let’s define information services. For the purposes of this blog post, information services refers to the industry that offers research, data, knowledge and instructional materials to customers. Think of publishing houses, data companies, research organizations, etc. Businesses like Reuters, Wiley, Oxford University Press or Clarivate.

Now, why are AI agents for information services all the rage?

AI agents for information services offer never-seen-before capabilities in leveraging structured, semi-structured and unstructured data toward a wide range of applications.

The ability of an AI agent to specialize can help it understand complex domain-specific information more effectively - which means output will be more accurate. Its ability to interact with the environment means that it can provide more relevant information to researchers.

Its scalability enables you to expand your business into new horizons, experimenting with new business models. With its integrations, you can get the output from the AI agent to a downstream dashboard, analytics engine or user-facing chatbot.

But how, you ask? 

How AI agents can be used in information services

To truly understand how AI agents work in information services, it’s best to explore some use cases.

Data retrieval and indexing

AI-powered intelligent document processing (IDP) can instantly extract key information stored in various formats, such as documents, databases, or even external sources like websites and blogs. This could be semi-structured data like PDFs, unstructured like scanned documents, rows and columns of spreadsheets, or HTML pages. 

What to do with that?

Data retrieval with agentic AI makes downstream applications possible. Let’s say you’re the legal department of an enterprise. Extracting data from all the contracts you’ve signed helps you: 

• Stay on top of renewals, compliance, penalties, etc. 

• Quickly pull up an obscure clause from any contract from any point in time

• Automate updates or remarks retroactively

• Dramatically reduce research time in case your legal team needs some information

Knowledge management

Once you have a record of all your data and metadata, you can deploy AI agents for better knowledge management. You can identify entities in your data, tag and categorize them, and document their relationships, creating comprehensive knowledge graphs that power future content operations.

This is what a multi-billion dollar analytics leader did with their unstructured data curated from company filings, patents, trademarks, domain registrations, market reports, surveys, etc.

Read their success story here

What to do with that?

Good knowledge management is like having a map of your content landscape. With it, you can take your business anywhere you please.

Recommendations: Identifying the connections between ideas and recommending the right content to the user. Great use case for news and research publications.

Curation: Identifying related content and curating collections for specific user groups. For instance, if you’re Getty Images or Reuters providing visual media as an offering, you can dynamically curate content based on user type, such as a fashion blog or a sports news publication.

Learning and development: AI agent as a personal coach for employees across organizations and industries. A good AI agent can observe user behavior — like trends in the way they write content or code - and coach them based on organisational knowledge.

Research support

A typical research publication, such as Taylor & Francis or Elsevier, has a large volume of content they own. They offer access to researchers based on a subscription. However, the access is often limited to your ability to ‘search’ and find the paper you need.

Agentic AI can dramatically change that. Content and data organizations can create AI agents that serve as the personal assistant of a researcher or human agent.

What to do with that?

Better search: Unlike keyword-based search, an AI agent can understand user needs in natural language and find relevant content more effectively. 

A Middle Eastern event management company did this for their sales team to accelerate the time to create proposals. 

Read case study here

Analysis: An AI agent can automatically collect data from surveys, social media updates, forum posts, etc. What’s more valuable is that it can also act as a sparring partner for researchers, spotting hidden trends and exploring deeper insights.

Summarization: AI agents can summarize papers for researchers to evaluate whether they are worth reading in detail. This significantly saves time and energy, accelerating research outcomes. 

Personalization: Over time, AI agents can also remember preferences and personalize the experience of interacting with the data platform itself, creating long-term loyalty.

Fact-checking: When specifically fine-tuned to your data, the risks of AI hallucination can be reduced significantly. With that, you can also create fact-checking, plagiarism-monitoring or citation-checking agents for subscription.

With the latest Deep Research offering from OpenAI, you can also make service requests for reports and studies with ChatGPT.

Analytics and dashboards

Data-driven decision-making is one of the most obvious use cases for any organization. AI agents for information services are uniquely poised to help enterprises with that.

Let’s say you’re The Weather Company, Market Watch or FIFA, insights from your data contribute to a huge part of your revenues. Agentic AI can expand the horizons of new revenue channels from your information.

What to do with that?

Dashboards: Offering subscriptions or charging for access to unique data in the form of custom-curated dashboards. For instance, Market Watch can create custom dashboards for users interested in specific stocks.

Recommendations: Using analytics to help users make decisions. The Weather Company can create custom offerings for farmers with recommendations based on the weather. It can also provide custom forecasts and reports about long-term climate change, pollution, or deforestation to inform policy decisions.

Predictions: Going beyond just recommendations, AI agents can perform data analysis and make predictions. In fact, it can also allow users to input their own variables and perform scenario planning. “What happens if Elon Musk sells Twitter?” can be a legit scenario you can plan for!

The possibilities of agentic AI are overshadowed only by the implementation challenges they pose. Anecdotally, over 80% of AI projects fail. Here are a few considerations to avoid that.

What to watch out for when bringing AI agents into information services

Implementing agentic AI means eliminating human oversight over a large number of processes. AI agents are designed to ‘autonomously’ complete tasks. Naturally, this poses significant risks.

Generalization vs. Specialization

Most agentic AI uses large language models, like ChatGPT, Google Gemini or Facebook’s LlaMa. These machine learning models are good at a wide range of routine tasks but struggle to perform specialized tasks.

A better alternative might be to use specialized small language models, which excel at 1-2 focused tasks. If necessary, you could create a multi-agent system to complete workflows.

Data bias

Typically, LLMs don’t offer access to their training data. Even if they do, evaluating the millions of data points for bias would be a fool’s errand. 

A better approach would be to use your own proprietary data (hopefully devoid of biases) to fine-tune chosen models. This way, you can also dramatically improve the accuracy of output.

The team at Tune AI are experts at exactly this. Speak to us today to see how you can fine-tune and deploy agentic AI for your enterprise use cases.

[Contact us now]

Privacy and security

Anything built with sensitive data carries enormous risks to data privacy and security. AI agents for information services are no different. One key aspect of building agentic AI is deciding which actions the bot can not or should not take for security reasons.

Create a robust governance model that can be applied to your AI agents. Assess your risks, set up guardrails around sensitive customer data, and train users on best practices. Read about Tune AI’s approach to governance here.

Reliability

AI models tend to hallucinate, i.e., generate inaccurate or completely fabricated content in response to prompts. It is also possible that the model pulls information from unreliable sources, especially if the search is run on the Internet in addition to your proprietary data.

Though model makers are working on it, and it has been reduced dramatically, the reliability of your AI agents for information services depends on accuracy. So, build slowly, prioritizing accuracy over performance in the beginning. Test the models repeatedly and optimize as you go along.

Implementation

Should you implement AI agents on the cloud or on-prem? What kind of infrastructure do you need? What tools can you securely integrate with? Who should be given access? What legal and ethical considerations apply?

Avoid unnecessary risks with implementation with a reliable partner with a track record of successful projects with AI agents for information services. 

Tune AI’s experts are uniquely skilled in implementing AI projects across industries, technologies and data types. Our work with some of the world’s largest data organizations has put us in good stead to serve your needs.

Talk to us about the possibilities of using AI agents for information services. Book a call today!