Is agentic AI sounding the dying knell for COBOL and mainframes?

Earlier this 12 months, pundits and publications ran with a brand new time period: SaaSpocalypse. This prognostication was based mostly on a number of developments in expertise by way of vibe coding, legacy code refactoring, autonomous brokers and Mannequin Context Protocol server capabilities.

With out query, main change is in retailer for the SaaS market. This more and more consists of transferring towards headless operations run by brokers reasonably than a conventional graphical consumer interface.

Even Manoj Swaminathan, a president and chief product officer at SAP, mentioned not too long ago that the seller is supporting this pattern by exposing its APIs. But when one factor is evident, CIOs aren’t going to interchange their main enterprise apps with one thing homegrown. I imagine that is for 2 causes:

With this mentioned, there are two applied sciences that CIOs have wished to interchange for years — and could possibly be changed by AI: the programming language COBOL (or widespread business-oriented language, designed in 1959) and that outdated trusty standby, the mainframe (which additionally dates again to the Fifties).

Associated:When AI brokers collaborate, who owns the outcome?

These applied sciences symbolize an earlier, on-premises-only period of computing. At the moment, corporations should preserve the expertise modernized and agentified — and the abilities wanted to maintain 75-year-old expertise working proceed to exit the enterprise.

This is not a brand new downside. COBOL code was so historic in 2000 that few folks knew the way to repair it to help going past 1999, as you might keep in mind. (Personally, I booked journey early to Hawaii for me and my younger household as a result of we nervous that journey techniques have been certain to fail as the brand new 12 months approached.)

COBOL continues to age, but it surely additionally stays entrenched (and dependable) for a lot of industries, together with airways, banking, insurance coverage, authorities, healthcare and retail. See a inexperienced display, and you have doubtless discovered a COBOL software powered by a mainframe.

Till earlier this month, I had not seen a pattern that would exchange COBOL and its workhorse, the mainframe — and I’ve labored in and written about enterprise expertise for 30 years.

True, low-code distributors have slowly eaten away at these historic traces of code for a number of years. However AI is having a novel impact the place brokers are turning the code for COBOL-based functions into an in depth specification for the appliance and its performance.

With this in hand, one other agent can rebuild the COBOL software utilizing a low-code, cloud-based growth stack, and one other can then conduct a regression check on the ensuing code.

Associated:Claudeforce and what comes subsequent for agentic AI

“Some of the thrilling areas is what we’re doing with COBOL packages,” Woodson Martin, CEO of OutSystems, advised me final month. “What we’re exhibiting enterprises is the power to develop necessities with an agent after which construct with brokers is right here — after which [they can] add agentic capabilities to what are COBOL packages.” OutSystems is a low-code platform centered on among the most intricate trade use circumstances.

The corporate launched Legacy Modernization Companies powered by AWS Rework in June to assist corporations migrate COBOL functions utilizing a multi-step strategy the place brokers scan code, construct dependency graphs, extract enterprise logic and be sure that a human stays within the loop.

Given what OutSystems is providing, I needed to ask: Is that this the dying of this legacy programming language?

“Sure,” Martin mentioned. “What we’re doing will result in the dying of COBOL and the dying or relegation of the mainframe.”

He pointed to the work his firm is doing with the insurance coverage trade, which vastly makes use of COBOL-based functions on mainframes. To manually migrate a COBOL software may take six months to a 12 months. OutSystems is utilizing agentic AI to chop migration time to only eight weeks. Now, Martin is hoping to get that time-frame right down to seven days.

Associated:Small language fashions promise financial savings. Here is what occurs later

“We wish it to be the cutting-edge, to be lifeless easy,” he mentioned. “That is the trail for a lot of to agentic AI.”

What does this all imply?

Agentic AI is clearly reworking the software program trade as a complete. It is going to change the worth proposition of SaaS functions, transferring them from techniques of file to techniques of motion.

To be clear, techniques of file will stay. Over time, the agentic AI pattern may have an effect on the rents that software program organizations extract from prospects. It is going to additionally allow tangible enterprise transformation as organizations pursue updates to their core enterprise capabilities and the best way they win.

The excellent news is that software program organizations are innovating quicker to help prospects and keep away from being left behind.

Lastly, the final vestiges of the early pc period is likely to be overcome by brokers in a position to extract and derive a specification and construct in low code a substitute for inexperienced display apps and computing.

Success right here will likely be a serious win for CIOs who must free assets for AI funding. And the power to use brokers to this new code will have an effect on corporations and their prospects as brokers service prospects for apps which are remade.

Do you assume agentic AI will mark the top of COBOL and mainframes? Tell us at [email protected].