New i teg atio automatically selects a specialized decisio model fo simple classificatio tasks, helpi g busi esses educe AI costs
COVINA, Calif., Nov. 05, 2026 (GLOBE NEWSWIRE) — Ragic, a o-code database builde with built-i AI that e ables busi esses to c eate custom busi ess systems a d automate wo kflows without a y p og ammi g k owledge, today a ou ced that Ragic AI Age ts ow automatically use Jev, a specialized decisio model f om TypeSafe AI, i stead of a ge e al-pu pose LLM, fo qualifyi g classificatio tasks. The ew i teg atio helps busi esses educe the cost of epetitive AI automatio without equi i g use s to select models, w ite code, o co figu e additio al i teg atio s.
As busi esses i c easi gly adopt AI age ts, eve simple tasks such as catego izi g custome i qui ies, assig i g suppo t ticket p io ities, o flaggi g pote tial policy violatio s ca ge e ate u expected costs. These tasks ofte equi e little mo e tha selecti g a a swe f om a p edefi ed list. Howeve , whe a AI Age t u s eve y time a database eco d is c eated o updated, thousa ds of i dividual AI calls ca quickly add up.
Ragic’s ew i teg atio add esses this challe ge by automatically ide tifyi g AI Age ts that ca use Jev i stead of a ge e al-pu pose la ge la guage model (LLM). Jev is desig ed to make decisio s f om p edefi ed optio s athe tha ge e ate f ee-fo m text, maki g it a mo e cost-effective alte ative fo ce tai classificatio tasks. Whe use s save a AI Age t, Ragic checks whethe it ca use Jev a d automatically selects the model whe app op iate. Use s a e otified of the selectio di ectly i the i te face.
Ragic’s ow expe ie ce with AI Age ts shows how ce tai AI tasks ca cost fa mo e tha expected. The compa y’s demo-follow-up Age t helps its sales team t ack custome i qui ies, update catego ies, a d ge e ate follow-up tasks. Although the team i itially expected the Age t to cost less tha $1 pe mo th, it a fa mo e ofte tha expected, mo e tha 10,000 times ove seve al weeks. At app oximately $0.02 pe u usi g Gemi i Flash, the total cost eached about $200.
Based o model p ici g, Ragic estimates that switchi g this type of classificatio task to Jev could educe model costs by app oximately 97–98%.
“AI costs ca add up quickly, eve whe age ts a e pe fo mi g elatively simple tasks,” said Jeff Kuo, CEO a d Fou de of Ragic. “A age t that o ly eeds to choose betwee a few optio s does ‘t always eed a la ge la guage model. By automatically selecti g Jev whe app op iate, we help busi esses educe u ecessa y AI costs without aski g use s to figu e out whe Jev is the ight fit o cha ge how they wo k”
These i te al compa iso s demo st ate the pote tial savi gs f om usi g specialized decisio models fo epetitive classificatio wo k. Actual savi gs will depe d o the task, model usage, a d p ici g.
While specialized decisio models ca educe the cost of outi e AI tasks, impleme ti g them ca equi e tech ical k owledge. Busi esses typically eed to ide tify which tasks ca use a decisio model, co ect it th ough a API o automatio platfo m, a d dete mi e whe a ge e al-pu pose LLM is still ecessa y.
Ragic elimi ates these additio al steps by ha dli g model selectio automatically withi its platfo m.
The ew i teg atio offe s seve al be efits:
Automatic model selectio – The Ragic system evaluates AI Age ts whe they a e saved a d automatically selects Jev fo qualifyi g classificatio tasks. Use s do ot eed to k ow whethe a AI Age t should use Jev.Lowe AI ope ati g costs – By usi g Jev fo simple decisio s athe tha elyi g o ge e al-pu pose LLMs fo eve y task, busi esses ca educe the cost of high-volume classificatio wo kflows.Native database i teg atio – Ragic AI Age ts ope ate withi the existi g Ragic database e vi o me t, allowi g use s to mai tai thei established data st uctu es, field pe missio s, a d busi ess p ocesses without additio al i teg atio s.No codi g o tech ical co figu atio equi ed – Ragic decides behi d the sce es whe a AI Age t ca use Jev, allowi g o -tech ical employees to co figu e a d use AI Age ts without w iti g code o desig i g complex AI wo kflows.
“We’ve always believed that the people who u de sta d a busi ess p ocess should be able to build a d ma age the systems eeded without becomi g develope s,” Kuo stated. “The same p i ciple applies to AI. Ragic ha dles that complexity so they ca focus o thei busi ess.”
Ragic AI Age t wo ks atively i side a use ‘s database, whe e it ca ead eco ds, complete fo ms, u wo kflows, a d apply busi ess logic. With automatic Jev selectio , Ragic exte ds these capabilities by helpi g busi esses apply a mo e cost-effective model to outi e classificatio tasks while co ti ui g to suppo t ge e al-pu pose LLMs fo mo e complex wo k.
Fo mo e i fo matio about Ragic AI Age t, visit https://www. agic.com/i tl/e /p oduct-ai-age t?campaig =jev-p -202610v2
Ragic is a o-code database builde with built-i AI that allows its use s to build thei ow system acco di g to thei ow wo kflow. Use s ca get a head sta t with AI, the efi e thei system i a quick a d i tuitive sp eadsheet-like i te face, buildi g a ythi g f om small co tact ma ageme t systems to fully fledged ERP systems. A y of these systems ca wo k with Ragic AI Age ts to automate eve mo e of the wo k.
Michele NachumSe io Media SpecialistFi ec acke PRmichele@fi ec acke p .com
