Try our new intelligent model routing solution, Arcee Conductor. Sign up today and get a $200 credit (~400M free tokens).
Product
Today we're excited to announce a new "Auto Tools" feature in our intelligent model routing solution, Arcee Conductor. Auto Tools routes your tool calling / function calling prompts to the optimal model–selecting only from models specialized in tool calling.
A few weeks ago, we launched our intelligent model routing platform, Arcee Conductor, which optimizes your use of AI models by sending each prompt to the ideal model for accuracy and price–selecting from a suite of our lightweight, specialized small language models (SLMs) as well as leading large language models (LLMs).
Traditional LLMs treat every query the same: they send each prompt to one monolithic model—usually at a high price point. But the reality is that each prompt is different. Arcee Conductor’s intelligent model routing ensures that every request is automatically dispatched to the AI model that's optimal for the task at hand, always balancing accuracy, specialization, and cost. The results? Smarter outputs and an average of 65% savings on total AI model usage.
Today, we’re excited to announce an addition to Arcee Conductor that transforms how tool calling is handled in production AI.
If you care about AI model performance and cost, and building truly useful, agentic applications, read on.
In the fast-evolving landscape of AI, the demands on AI models are increasing every day. We’re no longer in the early days of AI, when many businesses were satisfied with– or even blown away by–the conversational abilities of LLMs.
Now, both enterprises and developers are upping the game. They want to work with models that provide world-class accuracy and cost—and that’s exactly what we give them with Arcee Conductor.
But intelligent model routing is only half the story.
The next leap is tool calling, which allows the AI models to integrate seamlessly with external systems and APIs to power dynamic agentic workflows.
As organizations lean heavily into AI-powered agents for document analysis, search, workflow automation, and more, tool calling enables your AI models to actually take action—accessing live APIs, extracting up-to-date information, and orchestrating real-world outcomes.
Tool calling (or function calling) means that an AI model can trigger external tools, APIs, or custom functions as part of its reasoning process. It’s the essential capability that transforms static chatbot LLMs into dynamic, action-oriented digital assistants. With tool calling, your AI model can:
The bottom line: tool calling is the future of AI model interaction. If your AI models can’t trigger functions or hit APIs, you’re missing out on a huge leap in productivity and value. That’s exactly why we added the new Auto Tools mode to Conductor.
Auto Tools is Arcee Conductor’s specialized mode for any prompt involving tool/function calling. Instead of relying on a single model, Conductor intelligently routes your tool calling prompts to what it evaluates to be the optimal model from this list:
No more worrying about which model is right for your function calling prompt. Let the model router handle the complexity—delivering optimal performance every time, at the lowest cost.
For general use of Arcee Conductor (Auto mode), you can work directly in our UI, or via the Arcee AI API. Due to the nature of tool calling–which invokes external tools and systems, and tends to involve custom code–the new Auto Tools mode works only via the API.
Below is an example of how to use Conductor’s Auto Tools mode for a common business use case: generating a comprehensive market research report, including the most up-to-date financial information via the Yahoo Finance API.
You can see the full prompt in this notebook by our Chief Evangelist Julien Simon.
As part of the prompt, we define what we want the report to include:
We instruct the AI models to take into account the information we obtain via tool calling (from the Yahoo Finance API) as it drafts the report. Here is the list of specifications we provide:
1. Always start with analyzing the tool result if available
2. Structure your response clearly using markdown
3. Be detailed and thorough
4. Include relevant context from your knowledge
5. Support claims with data and sources when possible
6. Stay neutral and objective.
Finally, we break down the format in which we want the report delivered–specifying the following as required sections:
In the tool calling function of our prompt (“call-tools”), we set the “model” to “auto-tool”.
“auto-tool” selects the Arcee SLM “Caller-Large” for the tool calling, and the model returns the function call we should invoke to answer the query. Once we’ve invoked it and retrieved its response, we invoke Conductor again in “auto” mode to generate the report. In this particular case, Conductor selects the SLM Arcee-Blitz (a cost-efficient, highly-capable SLM that excels at report writing).
The final output is a comprehensive report, complete with the most current information thanks to the tool calling function:
REPORT GENERATED BY OUR PROMPT
Tool information: I have used the tool and it provides the CEO of General Motors.
Direct Answer: The CEO of General Motors is Ms. Mary T. Barra. Her full job title is Chairman & CEO.
Analysis: Mary T. Barra has been the Chairman and CEO of General Motors since 2014. She is the first female CEO in the history of a major global automaker. Under her leadership, General Motors has focused on transforming the company to a more sustainable and technology-driven future, with significant investments in electric vehicles, autonomous driving, and connected car technologies. Barra has also played a crucial role in navigating the company through challenges, including the global semiconductor shortage and the transition to a post-COVID-19 market environment.
Barra's background includes a 38-year tenure at General Motors, where she held roles such as Vice President of Global Manufacturing and Executive Vice President of Product Development, Purchasing, and Supply Chain. Her extensive experience in the automotive industry, particularly in manufacturing and operations, has been critical in driving cost efficiency and quality improvements across the organization.
Resources
General Motors top-level website: GM.com
Yahoo Finance page: Yahoo Finance GM
SEC filings URL: GM SEC Filings
Google search based on the user question: CEO of General Motors Google Search
In this example, the total cost of the prompt when routed to Caller-Large + Blitz is $0.0010, vs. a cost of $0.0139 when routed to GPT-4o.
In other words, when Conductor routes this prompt to the SLMs, you pay just 7% of what you would pay if you routed it to the LLM. And you can judge the output for yourself! We are perfectly satisfied with the overall quality of the report.
Now imagine you're a financial analyst at a Fortune 100 bank, with teams running reports like this hundreds of times per day. We're talking savings of hundreds of thousands of dollars or more, on your total AI usage.
Using AI models to power your tool calling will streamline processes and boost productivity; using AI models intelligently routed by Arcee Conductor can bring that tool calling to an even higher level of accuracy and cost-efficiency, ensuring that each prompt is processed by the model that will get you the best result at the lowest price.
Here are just some of the use cases where you can integrate AI-powered tool calling into your business:
Data Analysis and visualization - LLMs calling tools to analyze large datasets, generate visualizations, and create interactive dashboards for business intelligence.
Automated customer support - Calling CRM systems, knowledge bases, and ticketing tools to provide comprehensive customer support and issue resolution.
Scheduling and calendar management - Integrating with calendar tools to schedule meetings, manage appointments, and optimize time allocation.
Document generation and management - Calling document creation tools to draft contracts, reports, or proposals based on specific business requirements.
Email and communication management - Integrating with email systems and communication platforms to draft, prioritize, and respond to business communications.
Project management assistance - Calling project management tools to update tasks, track progress, and generate status reports.
E-commerce and inventory management - Integrating with inventory systems, product databases, and order management tools to optimize e-commerce operations.
Financial transaction processing - Calling accounting software, payment processors, and financial management tools to handle billing and financial transactions.
Sales pipeline management - Integrating with CRM tools to update customer information, track sales opportunities, and forecast revenue.
Social media management - Calling social media platforms and analytics tools to schedule posts, monitor engagement, and analyze campaign performance.
If you’re building agentic apps, document automation, or LLM-powered chatbots that need to interact with real systems, Arcee Conductor’s Auto Tools can work for you:
Sign up for Arcee Conductor today and get $20 in free credits. And let us know what you think of Auto Tools mode–drop us a note on X or over on Linkedin.
Manual claims are slow and costly. Discover how enterprise leaders can implement automated insurance claims processing effectively with this guide on application and success measurement.
The first quarter of 2024 isn’t over yet, but we’ve already had epic developments at Arcee – with TechCrunch announcing our seed round, and our decision to merge with mergekit.
Discover how enterprise automation can revolutionize your workflows, boost efficiency, and reduce costs. Learn about the latest strategies and tools, with real-world examples.
Subscribe to our newsletter
.svg)
