The race to harness artificial intelligence to create data products and software for financial institutions is at the heart of a collaboration between consultancy Synechron and technology developer Cognition.
New York-headquartered Synechron, which has longstanding expertise in providing software solutions that financial organisations use to transform their operations, has embedded Cognition’s Devin agentic engineering platform to give banks and other institutions the tools to build their own automated data products.The partnership makes use of Synechron’s accelerator tools that enable users to modernise their legacy systems with the latest technology to create software at speed. By applying agentic AI, new software can be built without granular understanding of coding.
David Sewell, Chief Technology Officer at Synechron, said the collaboration puts the power of software creation into the hands of the people that most need and use it, without having to work through engineering teams, a process that will reduce the time it takes to put their programmes into production.
“We’re heading towards a paradigm shift in how software is created,” UK-based Sewell told Data Management Insight. “We’re seeing a different way of building software that allows for very rapid prototyping and business feedback, which then leads to fast engineering development of a more hardened business solution.”
Agents Working While You Sleep
Agentic automation is the latest frontier in AI’s development, enabling large language models to run workflows and processes without the need for continual human input. Data and technology companies like Snowflake and Microsoft have been busy building the capabilities for financial organisations to create agents.
Synechron, which was formed in 2001 and now operates out of 60 offices in 20 countries, teamed with San Francisco-based Cognition because the companies recognised each others’ complementary strengths, said Sewell.
Its customers have already been using Devin while Synechron had been testing its capabilities in its Synechron Labs, where the company builds internal products, tools and accelerators.
“Combining the leading agentic-based engineering system with our deep knowledge of financial services brings a very powerful capability to our clients,” Sewell said. “Once you realise that the power of the two of us together is so much greater than the individual, then it makes absolute sense to work together.”
A Natural Language Future, With Guardrails
In its initial iteration, the users that create the software are provided with structural guidance via an interface. Sewell stressed that this would be needed in particular for data access.
Engineers will still be required to help validate the resultant software and also to oversee the guardrails and governance provisions that will protect the data and IP, as well as ensure they are compliant with regulations and internal controls.
“Cognition and the Devin platform are not just producing software and deploying it to production directly – you still have to go through all of the change controls, all of the existing processes, all the pipelines, the data access controls, the identity and access management that you need to get access to data, those sorts of things,” Sewell said. “The method of getting solutions productionised is exactly the same. It’s producing the software in the first place that’s being accelerated.”
The eventual aim of the collaboration is to allow users to describe their business requirements in plain English, while the Devin platform, integrated with an enterprise’s organisational knowledge, runbooks and business processes, learns about the specific environment. This contextual understanding will enable it to interpret requests accurately and generate a “reasonable first shot at a product or a solution”, said Sewell.“The interface that the users tend to work within – and this is one of the real powers that Cognition brings to play – is an easy user experience to get access to your data sources, to get playbooks and skills that you’ve configured and shared within the environment,” he said.
“But ultimately, you’re trying to get to the point where somebody that understands the business requirements can type the business requirements in plain English, and the system will understand that.”
Subscribe to our newsletter
