Engages thousands of suppliers at once in 40+ languages
Cuts purchase request review times to 30 seconds with Gemini
Secures up to 15% savings on negotiated purchase requests
Achieves highest LLM accuracy at lowest cost with Gemini
Pactum AI uses Gemini to enable autonomous procurement processes for its enterprise clients.
Pactum AI uses Gemini to enable autonomous procurement processes for its enterprise clients.
Solving purchasing bottlenecks with Gemini and Google Cloud
Managing procurement for a global enterprise is highly complex. Every purchase request must be checked against policy, contracts, and approved suppliers. When performed manually, this mandatory work is slow and time consuming. Pactum AI set out to solve this challenge with its autonomous procurement platform. By providing companies with a workforce of AI agents, Pactum AI can conduct negotiations and handle routine purchasing requests at scale, doing the work of an entire team in a fraction of the time.
Early in its growth, Pactum AI hosted its platform on a third-party cloud provider, but the strict data security requirements of its major clients promptly led the company to migrate to Google Cloud to establish a secure, enterprise-grade foundation for sensitive supplier data.
As Pactum AI set about building its AI capabilities, it tested self-hosted, open-source large language models (LLMs). This, however, proved expensive, as the company had to pay for computing power even when the models were idle. Relying on these older LLMs also posed a business risk. Pactum AI frequently needs to use LLMs to extract sensitive data from unformatted documents, such as PDF invoices. If a model misreads a document and extracts the wrong email address, for example, the system might send confidential data to the wrong person. The company needed a smart, reliable LLM that could read these documents accurately, without driving up costs.
In our internal benchmarks, Gemini consistently delivers the most accurate results, at a much lower cost than the alternatives. Being able to source both our cloud infrastructure and our AI models from Google Cloud, meanwhile, improves data security and compliance for our clients and simplifies our development processes.
Taivo Pungas
CIO, Pactum AI
Pactum AI turned to Gemini models on Google Cloud. Gemini provided highly accurate answers, while operating on a flexible pay-per-use model, removing the need for dedicated hardware. As a participant in the Google for Startups Cloud Program, Pactum AI benefited from credits to put towards its infrastructure as well as both commercial and technical support from the Google Cloud team, helping the company make the most of its Gemini usage.
“In our internal benchmarks, Gemini consistently delivers the most accurate results, at a much lower cost than the alternatives,” explains Taivo Pungas, CIO at Pactum AI. “Being able to source both our cloud infrastructure and our AI models from Google Cloud, meanwhile, improves data security and compliance for our clients and simplifies our development processes.”
Automating complex purchasing tasks with Gemini platform
Pactum AI runs its core platform on Google Cloud. Using Gemini platform and the Agent Development Kit (ADK), the engineering team builds and manages an ecosystem of specialized agents that plug directly into a client's existing buying processes. These diverse agents work together to complete entire workflows for Pactum AI’s customers, across three broad areas of procurement.
Because Gemini models are so good at reasoning, they understand exactly what the user is trying to achieve. That understanding can be the difference between our customers needing 20 attempts to get the right answer and only needing two, making the process much more efficient.
Taivo Pungas
CIO, Pactum AI
First, in Pactum AI’s supplier interaction engine, custom AI agents, running on Gemini, reach out to thousands of suppliers to negotiate terms or update certifications. When a supplier replies, an agent uses Gemini to read the email, understand what the supplier wants, figure out if a human needs to step in, and translate the conversation to relevant local languages. While Gemini handles this natural language communication, Pactum AI uses strict programming rules to dictate the financial terms of any deal, ensuring clients are protected from unexpected outcomes.
Second, Pactum AI has built a range of alignment agents that automate back-office operations. For example, if an employee at one of Pactum AI’s enterprise clients asks to buy a new laptop, a specific alignment agent uses Gemini to instantly read the request, check it against company policies, and approve or reject it.
Third, Pactum AI’s materials intelligence agents help human buyers make better decisions. If a supplier wants to raise their prices due to changing material costs, for example, a decision-support agent uses Gemini to read complex parts lists and compare them against market data, helping buyers figure out if the price increase is fair.
To process these high-volume tasks quickly and affordably, Pactum AI uses Gemini Flash and Flash-Lite models, ensuring reliable performance at scale.
“Because Gemini models are so good at reasoning, they understand exactly what the user is trying to achieve,” notes Pungas. “That understanding can be the difference between our customers needing 20 attempts to get the right answer and only needing two, making the process much more efficient.”
Cutting purchase review times to 30 seconds with Google Cloud
Pactum AI’s autonomous procurement platform has transformed how enterprise clients operate. The alignment agent, for instance, cuts the time it takes to review a routine purchase request from up to three days to just 30 seconds. This instant response means employees no longer wait around for approvals, making the business run more efficiently.
By automating these everyday processes, human buyers have more time to focus on strategy. Meanwhile, the platform’s agents can comfortably communicate with thousands of suppliers at once, ensuring no supplier relationships are neglected. This combination has helped Pactum's clients secure up to 15% savings on purchase requests that previously went straight to purchase order without negotiation.
Using managed cloud services from Google Cloud also helps to keep Pactum AI’s cloud spend under control. “Google Cloud allows us to scale our usage up and down, meaning we don’t need to buy and maintain dedicated servers, and we only pay for the compute power we use,” Pungas explains. “This flexible cloud foundation ensures our platform remains fast and responsive, even during periods of high transaction volume.”
As Pactum AI grows, the company plans to use Gemini to scale its agentic workforce to serve more customers across new procurement challenges. The engineering team is developing general empowered agents that can integrate into multiple parts of the purchasing process, automatically reading complex legal contracts and tracking changing compliance rules to keep daily operations in sync with company policies.
Google Cloud allows us to scale our usage up and down, meaning we don’t need to buy and maintain dedicated servers, and we only pay for the compute power we use. This flexible cloud foundation ensures our platform remains fast and responsive, even during periods of high transaction volume.
Taivo Pungas
CIO, Pactum AI
“As we expand our agentic workforce, we want to build highly dynamic workflows that adapt to any procurement scenario,” concludes Pungas. “Using Gemini to empower our agents with deep document intelligence will allow us to tackle increasingly complex challenges and reshape how global enterprises manage their suppliers.”
Pactum AI is an autonomous procurement platform that helps enterprise companies automate complex procurement processes and execute supplier negotiations at scale.
Industries: Technology, Startup
Location: Estonia
Products: Agent Development Kit (ADK), Gemini, Gemini platform, Gemini Flash, Gemini Flash-Lite, Google Cloud
