Malt’s AI Search tool understands what companies are looking for in a freelancer and creates matches that go beyond keywords. Malt is a platform that connects companies and freelancers for knowledge-intensive projects, including many IT and technology assignments.
“Belgium is a mature country in terms of freelancers,” begins Claire Lebarz, Chief Technology Officer at Malt. Although the supply is very broad, companies struggle to clearly formulate their needs and select the right person from thousands of profiles. That is why Malt is committed to AI-driven matching.
“With AI Search, we help companies find not just someone with the right skills, but the right freelancer who fits the context, the project and the organization,” she says. “For freelancers, this means less time spent on unsuitable requests and more assignments that match their expertise and ambitions.”
AI searches and refines
To simplify the search for freelancers, the company developed ‘AI Search’ on top of its existing sourcing platform. Companies can now use AI to refine their search for freelancers for a variety of projects.
For example, companies can ask AI Search: ‘we are looking for someone for X, in context Y’, or upload a PDF with a job description. “About 90 percent of customers start their search with a prompt, only a small proportion use a PDF for this,” he knows. That search is then translated into “Malt language”. This includes concrete skills, seniority level, type of role and context.
read also
How AI simplifies the search for freelancers, and vice versa
The result is a short list of possible freelancers who are not only selected on the basis of a few keywords, but fit within the description.
No black box
Malt does not want to be a “black box”, but is working on conversational AI that asks targeted questions and helps clients write better project descriptions, while the user retains control. “As soon as we have converted the search query into Malt language, customers can still make corrections,” says Jactthar. “We focus very strongly on the experience of customers. That way they understand what is happening,” he adds.
Own language models
Malt consciously chooses to use its own developed language models instead of the large AI models such as ChatGPT. “We have a team of machine learning engineers,” begins Lebarz. “They developed AI Search based on small language models trained on millions of data points from more than ten years of Malt data. In this way, they were able to include the context of the talent, the job description and project description in our language models.”
In addition, proprietary language models ensure that “we have more control over performance and limit the tool’s carbon footprint,” she adds.
Time-saving
Moreover, it appears that the AI tool saves companies a lot of time. “With larger corporate clients, we see the biggest leap in autonomy: previously, about 20 percent felt confident enough to place an assignment on the platform without help, today that is about 60 percent,” says Jactthar. “With SMEs, that process already runs smoothly independently.”
“Thanks to AI, customers have become much more autonomous: previously, only about 20 percent had enough confidence to put an assignment on the platform without help, now that is about 60 percent,” Jactthar knows.
read also
AI Saves Belgian Employees 108 Minutes per Workday
“Entering their needs is also much faster: with AI it takes an average of four minutes, compared to about 20 minutes via the classic search method where they have to write, search and filter themselves,” he adds.
Relevance
In addition, the entire process is accelerated: where customers previously only received the first application from a freelancer after 16 to 24 hours, today about 80 percent of customers receive a response within an hour. Not only faster, but also more relevant: most proposals come from freelancers whose expertise closely matches the assignment.
read also
AI adoption is increasing faster than awareness of security
Malt’s AI Search tool first tries to understand what the customer is really looking for, in order to find the best possible profiles. Not on the basis of individual keywords, but by interpreting the entire project in terms of content. “We have the ambition to grow into an AI-first company that supports not only matching, but all aspects of working with freelancers with AI,” concludes Lebarz.