Ai Use Cases That Actually Fix Engineering Bottlenecks

This is an excerpt from Chapter 3 of “AI for nan Enterprise: The Playbook for Developing and Scaling Your AI Strategy,” a caller ebook by acclaimed tech journalist Jennifer Riggins and sponsored by Red Hat and Intel.

From nan advantages of utilizing nan “two-speed” AI finance model, to measuring nan existent effect of AI, this free book, now disposable for download, helps endeavor leaders create an AI strategy to unlock productivity gains, lick antecedently intolerable problems and summation a existent competitory edge.

If penning codification was ne'er nan bottleneck and codification procreation is already wide adopted, past organizations request to look elsewhere for caller AI usage cases.

There will beryllium cases of one-size-fits-all AI. SAP’s astir celebrated AI usage lawsuit by acold is scanning and processing receipts — thing galore labor do. But much likely, AI will lick for smaller pains, often circumstantial to roles aliases functions. Keep things thin and elemental by starting pinch early wins that could stably turn successful nan future. Then, erstwhile a usage lawsuit is proven, you tin accommodate it for different teams.

While location are plentifulness of nontechnology problems to solve, nan first AI usage cases thin to beryllium incubated wrong tech teams, Hannah Foxwell, laminitis of AI for nan Rest of Us, said. It feels safer to neglect successful nan tech section because nan soul information utilized by tech teams whitethorn not person arsenic overmuch effect arsenic finance, quality resources and customer data. AI could person a greater effect successful those areas, but they besides person a overmuch higher risk.

Documentation and Technical Debt

Developers are accordant successful explaining what astir slows them down: insufficient, out-of-date archiving and method debt. Neither of these are issues astir devs want to hole themselves. Both besides worsen pinch a accelerated summation successful codification — and some are usage cases wherever AI tin radiance because it is awesome astatine having conversations and explaining complexity.

In fact, nan biggest AI triumph uncovered successful nan 2024 “DORA” study recovered that a 25% summation successful AI take triggered a 7.5% summation successful archiving quality. Take this proven usage lawsuit companywide by adding a chatbot overlay for soul information, ownership and processes.

Rachel Laycock, CTO of Thoughtworks, finds that bequest modernization is still nan apical situation facing astir enterprises. “It’s not clear erstwhile we will get to a settled abstraction of nan apical 3 aliases 5 devices and models group use,” she said.

Additionally, she said, group are astir often building “greenfield” apps, which is easy compared to building features that person to beryllium integrated into bequest applications. The market, she continued, is “focusing excessively overmuch connected ratio of producing code, which isn’t really nan problem.”

Code Review

Another bottleneck that increases pinch improvement velocity is codification review, which leaves developers waiting longer for feedback and increases cognitive load.

“Previously, value was derived from having quality eyes astatine 2 steps — nan developer who wrote it and nan PR [pull request] reviewer,” Foxwell said. “Now nan developer who ‘wrote’ it whitethorn not person publication it. It whitethorn person been written by AI. It looks good to ship, and nan personification reviewing it is faced pinch a upland of codification to reappraisal each astatine once.”

AI is besides useful here, observed Nathen Harvey, DORA lead and merchandise head astatine Google Cloud. By having nan correct specifications, archiving and guardrails successful place, you tin supply rapid, automated feedback to developers utilizing AI arsenic a first-round PR reviewer. Then AI tin delegate nan PR to nan adjacent pinch nan readiness and nan knowledge to reappraisal it. This reappraisal besides takes little clip and is much constructive because this quality successful nan loop tin beryllium assured that nan PR adheres to established organizational standards.

Unlocking Insight successful Data

Some organizations, for illustration successful financial services aliases governments, person 100 years’ worthy of information that they tin yet unlock pinch AI. However, that process whitethorn not beryllium an easy one, particularly wrong budgetary, regulatory, privateness and information guidelines.

When reasoning astir AI usage cases, Red Hat Senior Product Marketing Manager Marty Wesley remarked: “The easiest ones are successful financial services, because there’s conscionable truthful overmuch data, and they tin usage nan AI to thief cod a batch of that accusation and make decisions connected it.” Consider nan often tedious, in-person, days-long acquisition of opening a commercialized account, which requires companies to taxable each kinds of system and unstructured information crossed different documents.

AI tin publication nan existing forms and get a consciousness of nan business, allowing nan slope to move nan full process online successful a overmuch shorter clip frame. And this isn’t justifying unit layoffs astatine these “accidental tech companies,” but alternatively restructuring for worth by reassigning unit to different areas that person issues, problems, aliases conscionable request much help, Wesley recommended.

Of course, mounting up an relationship is simply a repeatable AI usage lawsuit communal to each businesses, from car security to assemblage applications.

Another AI-driven usage lawsuit is scanning unstructured information successful images and videos much quickly than humans could ever imagine. For example, Boston Children’s Hospital and a proscription institution are utilizing Red Hat exertion to uncover patterns of disease and inspect railway tracks utilizing drones, respectively.

Boston Children’s Hospital says its AI radiology devices “are arsenic bully arsenic nan quality radiologists, and they’re going to get better,” Wesley said.

Safe Experimentation

If you deliberation you had instrumentality sprawl before, hold until you adhd successful AI. No statement tin spend to adopt each instrumentality retired there, but they besides can’t spend to miss retired connected nan adjacent AI productivity gain. Enterprises request a robust civilization of experimentation supported by clear permissions, processes and guidelines.

A bully first measurement is encouraging teams to A/B trial different products and connection feedback, for illustration package improvement institution Devexperts did when testing and comparing codification generator devices Copilot and Cursor, past sharing trial methods and results.

At this velocity of change, your main AI agency will apt go a halfway of excellence for AI experimentation. An AI aviator aliases impervious of conception (POC) is “your chance to make judge a instrumentality will really present results for your organization: improving developer experience, accelerating improvement and ensuring your engineering practices still align pinch your foundational meaning of excellence,” explained DX CTO Laura Tacho.

To publication more, download “AI for nan Enterprise: The Playbook for Developing and Scaling Your AI Strategy” today!

Group Created pinch Sketch.