AI & Automation

Startup Ideas

• 2026-08-10

  Nobody Can Say Whether the Engineer Burning $50K a Month Is Waste

  Even companies that break AI spend down by employee cannot see whether that spend came back as output, so there is no basis for deciding whether to block or back the engineer spending $50K a month.

• 2026-08-06

  Trading In Your Old Game Paid for the Next One and That Loop Is Ending

  A disc price always contained money you could recover later by trading it in, and a digital license contains none of it, so the same sticker price quietly costs more.

• 2026-08-03

  Sony Secured Its RAM. Who Is Watching Your BOM?

  DRAM and NAND prices jump by double digits every quarter, yet only large companies can lock in supply with advance purchases, while small manufacturers learn about price moves late and have no tools to respond.

• 2026-07-27

  Why Does Everything a Forward Deployed Engineer Built Evaporate Every Time

  What a forward deployed engineer builds inside a customer's environment is locked in by NDAs and data residency rules, so the third deployment costs exactly as much time as the first.

• 2026-07-27

  One Checkbox Saying a Game Used AI Tells You Nothing

  Steam's AI disclosure is effectively one on/off box, so a game where AI is the mechanic, a game selling generated art assets, and a game that used AI only for localization all carry the identical badge on the same shelf.

• 2026-07-26

  AI Writes the Code, So Why Is the Back Office Still Manual?

  Computer-use agents finish short tasks but complete only about one in five real office workflows end to end, while vendors bill on savings delivered. Nobody measures what was actually saved.

• 2026-07-22

  Where You Cut the Episode Decides the Revenue

  In one to two minute vertical dramas, where you cut each episode and which episode carries the paywall drives the revenue, yet both decisions are made on an editor's instinct and never travel to the next production.

• 2026-07-20

  Agents Have Started Spending Money on Their Own, and No Control Layer Is Watching

  Rails are giving agents the power to pay, but nothing controls a runaway charge in real time or tracks where the money went. One malfunction or prompt injection turns straight into an invoice.

• 2026-07-19

  You Can Buy GPUs but Not Electrons: The Power Bottleneck Is a Product Gap

  The binding constraint of AI datacenters has shifted from compute to power, yet the software layer that meters electricity, procures it, and moves workloads around it barely exists.

• 2026-07-13

  Nobody Actually Knows if AI Improved Output: Measure the Work, Not the Usage

  Companies mandate AI use and count seats and prompt volume like a KPI, but nobody honestly measures whether that usage actually lifted an individual's or a team's output.

• 2026-07-08

  Voice-API Bills Eat Your Margin, Can On-Device Speech Flip That

  Audiobook, accessibility-reader, language-learning, and dubbing products pay cloud voice APIs by the character or minute, so cost scales linearly with usage and margins get worse as you grow.

• 2026-07-07

  A 7MB Model Runs in the Browser CPU in 5ms, the Products You Can Build on Server-Free Inference

  Even basic AI features like semantic search or intent classification today require sending every request to a server API, which adds latency, piles up token cost, ships user data off the device, and stops working entirely offline.

• 2026-07-06

  Your Demo Video Went Stale on Yesterday's Deploy, the Missing Demo-Media Refresh Layer

  Landing pages, docs, app-store screenshots, and onboarding videos all show the product UI, yet the product ships weekly, so within days they display old buttons and features that no longer exist. No team clearly owns keeping them fresh, so they rot until a customer complains or a rebrand forces a batch redo.

• 2026-07-03

  We're Handing Agents the Whole Keyring, the Missing Credential Broker

  For a coding agent to push code or call an API today, someone hands it full human credentials, and one prompt injection or leaked debug log is all it takes for those keys to walk.

• 2026-07-02

  AI Uses Your Voice, But There's No Front Door to Permit It and Get Paid

  Creators' voices and faces are trained on and synthesized without permission, but there's no standard place to register, track, and charge for that use, so they defend themselves one lawsuit or viral callout at a time.

• 2026-07-01

  When Agents Improve Themselves, Nothing Decides Whether You Can Trust the Change

  A self-improving agent claims it's better than yesterday, but the team has no way to verify the claim. There's no gate to tell whether a new prompt lifted task A by 5% while dropping task B by 30%, or whether the model's self-scored 'pass' is a real gain or just reward hacking, learning the grading rubric instead of the skill. The space between deploy and rollback is wide open.

• 2026-06-30

  Arena Became a $100M Company, But There's No Tool to Build Evals for Your Own Product

  Public leaderboards rank models on an average. Founders need to know which model and prompt answer best for the questions their users actually ask inside their product, and the two often disagree. The Arena's #1 can be your domain's #3. But with no tool to build product-fit evals, most teams paste a few outputs into a spreadsheet and pick by gut, redoing the work from scratch every time they swap models.

• 2026-06-28

  No One Measures Where AI Actually Helps vs Hurts: Task-Level Augmentation Analytics

  AI adoption is decided company-wide, but its effect splits task by task. The same model is faster than humans at one step and quietly wrecks quality at the next by missing subtle exceptions. Yet most firms only watch pre/post cost savings, they can't separate where AI adds value from where it subtracts it. Ford never measured that difference, cut its veterans, and paid for it in recalls and rehiring.

• 2026-06-27

  Rather Than Wait Five Years for a Grid Connection, Make Your Load Flexible: A Control & Verification Layer for AI Datacenters

  What stops a developer from standing up an AI datacenter isn't generation capacity, it's the grid interconnection queue. Utilities study new loads on the assumption they run at peak output all year, and in major grids that line runs two to five years. A 'flexible interconnection' path that lets you skip the queue by giving up a sliver of load is opening up, but there's no software to actually shed AI training and inference on a single signal without breaking SLAs, so the datacenter can't even make the promise.

• 2026-06-26

  Letting an Agent Run for a Day: Operations Infra for Long-Horizon Coding Agents

  Let an agent run for more than an hour and every failure mode invisible in short tasks surfaces at once. Context gets compacted and the agent drifts from its original spec. Token cost piles up linearly with no governor. A 25-hour run can quietly collapse at 3am and nobody notices until morning. One practitioner warns that long-running mode 'consumes truly absurd amounts of tokens,' and five-hour plan limits expire before the work is done. You can launch a long-horizon agent, but there's no instrument panel to see whether it's running well, stop it, or revive it.

• 2026-06-25

  Agents Can't Find Their Own Tools, A Runtime Tool-Discovery Layer

  Agent capabilities are tied to manual pre-installation. You hardcode a tool URL into config, and the fallback, dumping every tool description into the LLM's context, hits token budgets and blurs tool disambiguation. There's no way to discover a tool you've never met while the task is running.

• 2026-06-25

  Employees Are Burning the Whole Token Budget: Build AI Spend Governance

  Uber blew its entire 2026 AI budget in four months and capped employees at $1,500 per coding tool per month. Walmart throttled tokens on its internal vibe-coding platform; Amazon told staff to stop using AI 'for the sake of using AI.' Mid-size orgs have no tool to see, cap, or reroute per-employee and per-team LLM spend to cheaper models.

• 2026-06-25

  CLIs Built for Humans Don't Work for Agents, The Agent-Native Tooling Market

  Decades of CLIs were built assuming a human watches the screen and presses keys. Progress bars, interactive prompts, color-coded tables, errors dumped into stdout alongside data, friendly to people, pure noise to agents. An agent can't press a key at a stalled prompt, and it burns tokens trying to parse output it can't parse.

• 2026-06-24

  Agents Have Started Learning Your Company, The Audit Layer That Proves 'What It Knows'

  Persistent agents like Claude Tag learn from every channel they're granted access to and build up organizational memory. But enterprises have no way to audit exactly what an agent knows, which conversation it learned it from, or whether that memory leaks from one department to another.

• 2026-05-30

  Enterprise IT Agents Miss More Than Half: The Empty Seat Called the Verification Layer

  Enterprises are told to hand IT operations to autonomous agents, yet frontier models get more than half of SRE incident-resolution tasks wrong. Worse, more investigation turns produce more false positives, not better answers.

• 2026-05-29

  Why Enterprise AI Costs Spiral Out of Control, And What's Missing Is Context

  Companies adopted AI but costs exceed projections and accuracy disappoints. Root cause: AI operates without enterprise context, generating redundant API calls and repeated queries to piece together answers that should be obvious.

• 2026-05-26

  1M-Token Context Exists: But No App Is Built for It Yet

  Even with 1M-token LLMs available, existing apps are architected for short contexts, making real agent work, full codebase analysis, long contract review, cross-session memory, unreliable or impossible.

• 2026-05-23

  AI Startup ARR Is Inflated 3-5x: Who Fills the Gap?

  AI startups routinely conflate CARR (Contracted ARR) with actual ARR, or include customers not yet onboarded, inflating metrics across the industry.

• 2026-05-23

  When AI Inference Cost Becomes a Routing Problem, Who Decides Where Each Request Goes?

  AI service teams spend 30-60% of revenue on inference but have no single dashboard showing which model, chip, or region is cheapest per request.

• 2026-05-22

  What If A/B Testing Cost Nothing? SimGym and the Era of VLM-Simulated Experiments

  LLM agents simulate A/B tests without real traffic, compressing e-commerce experiment cycles from weeks to under an hour

• 2026-05-10

  Foundry Geopolitical Risk Intelligence Platform

  Fabless semiconductor companies understand TSMC concentration risk but have no dedicated tool to quantify foundry geopolitical risk or simulate the cost of switching to Intel IFS, Samsung Foundry, or GlobalFoundries. Board-level supply chain decisions rely on consultant reports and spreadsheets.

• 2026-04-18

  Automatic Vendor Selection System to Solve Unpredictable Cloud Costs for Small and Medium Enterprises

  Small and medium enterprises face extreme difficulty predicting cloud and external service costs that fluctuate wildly based on usage. This accelerates cash burn, prevents focus on core product development, and leads to premature startup failures. Existing dashboard tools rely on manual analysis and fail to reflect Korean regulatory nuances or language-specific service characteristics.

• 2026-04-15

  Automatic Optimal Path Selection to Solve Surging Compute Costs for Enterprises

  In 2024 many Korean companies saw their compute bills from overseas providers increase 4-7 times. Existing tools are English-first, lack deep integration with domestic clouds and ignore data residency regulations. Development teams cannot forecast spend, budgets are exceeded, and CFOs lose control.

• 2026-04-12

  A Customized Shift Scheduling System to Solve Nurse Turnover in Regional Small and Medium Hospitals

  Regional small and medium hospitals suffer from high nurse turnover rates due to poor working conditions and irregular shift work compared to large hospitals. Unfair scheduling relying on manual work increases nurses' stress and reduces work efficiency.

• 2026-04-07

  An Automatic Building Condition Analysis and Repair Vendor Matching System to Solve Sudden Facility Breakdowns in Mid-sized Commercial Buildings

  Unlike large prime offices, mid-sized commercial buildings lack professional facility management teams and rely on manual records or intuition. Sudden breakdowns of HVAC, elevators, or plumbing cause massive tenant inconvenience and lead to significantly higher costs due to emergency repairs.

• 2026-04-06

  An Automatic Building Condition Analysis and Advance Notification System to Solve Frequent Facility Breakdowns in Small Commercial Buildings

  Unlike large buildings, small commercial buildings lack resident facility managers, meaning severe issues like leaks or HVAC failures are only noticed after they occur. This leads to emergency repair costs that are multiple times higher than normal, severe tenant dissatisfaction, and ultimately a decline in property value.

• 2026-04-03

  An Image-Based Multilingual Detail Page Automatic Generation Service to Solve Complex Overseas Sales Content Creation for Small Online Sellers

  Most small sellers lack the budget and design skills to create localized, engaging product pages for foreign markets. Simple translation tools fail to capture local consumer sentiment, and manually re-editing images for different regional marketplace standards is extremely time-consuming.

• 2026-04-01

  A Smart Task Allocation System to Solve Inefficient Routing Problems in Small Warehouses

  Most small logistics warehouses rely on the intuition of experienced workers to perform picking tasks. This leads to inefficient work routes and a sharp drop in productivity when new workers are introduced.

• 2026-03-30

  An Integrated Usage Analysis and Cost Optimization Service to Solve the Hidden Software Subscription Waste Problem for Small Businesses

  Most companies pay for numerous online tools individually by department, spending unnecessary monthly costs on accounts of former employees or unused services. There is a severe lack of visibility to track and manage this centrally.

• 2026-03-28

  An Automated Usage Tracking System Based on Payment History to Solve Unseen Software Subscription Cost Leaks for SMEs

  Most SMEs subscribe to individual software by department, but accounts of former employees or programs with overlapping functions are neglected, wasting massive amounts of money every month. Without a centralized management system, management cannot even grasp the exact scale of expenditure.

• 2026-03-26

  An Automatic Usage Analysis and Management System to Solve Unnecessary Software Subscription Waste for SMEs

  Most SMEs subscribe to dozens of software tools, but waste massive amounts monthly due to forgotten accounts of former employees or duplicate payments. Without a dedicated IT manager, manual tracking is virtually impossible.

• 2026-03-25

  An Automatic Call Voice-to-Text Conversion and Core Content Summarization System to Solve Consultation Delays and Record Omissions in Small and Medium Call Centers

  The structure where agents type while talking to customers severely degrades work efficiency. This leads to longer customer wait times and a vicious cycle where important complaints or promises are missed, resulting in secondary complaints.

• 2026-03-24

  A Data-Driven Purchasing Timing Recommendation System to Solve Surging Ingredient Costs for Restaurants

  Ingredient costs are the biggest variable in restaurant operations, but small owners have no way to foresee price spikes caused by extreme weather or supply chain issues. The reactive structure of buying at high prices when inventory runs out severely damages operating profits.

• 2026-03-23

  An Automatic Optimal Supplier Matching System to Solve Complex Ingredient Price Fluctuations for SME Franchises

  Purchasing managers at SME franchises suffer from severe inefficiency, having to compare prices daily with dozens of suppliers via phone or Excel. Frequent price changes cause them to miss optimal purchasing timing, leading directly to lower profit margins.

• 2026-03-22

  A Country-Specific Design and Translation Auto-Conversion System to Solve the Complex Overseas Promotional Material Creation Problem for Exporting SMEs

  Exporting SMEs often lack the budget and specialized personnel to create marketing materials that fit the cultural context and preferences of target countries, even if their products are competitive. Simple translation is insufficient to resonate with local consumers, and using the wrong colors or images can significantly damage brand trust.

• 2026-03-19

  A Conversation-Based Automatic Booking System to Solve Customer Management Issues for Single-Person Salons

  Single-person salon owners handle treatments, cleaning, and customer service alone, making it difficult to answer booking inquiries during treatments. This leads to missed bookings, customer churn, and significant revenue loss.

• 2026-03-18

  An Automated Photo-Based Sorting System to Solve Complex Return Inspection Delays for Small Online Malls

  While online shopping returns are increasing, small businesses must manually check the condition of each returned item. This delays refunds, increases labor costs, and frequently causes conflicts with consumers due to the difficulty of proving the exact condition.

• 2026-03-16

  An Automated Response and Scheduling System to Solve Complex Patient Booking Issues for Small Clinics

  Clinic receptionists handle dozens of simple calls daily regarding bookings, locations, and hours, degrading in-person patient service and increasing burnout. Missed appointments or delayed responses directly lead to patient churn and lost revenue.

• 2026-03-08

  An Automatic Bid Document Completion and Management System to Solve Complex Public Procurement Procedures for SMEs

  Supplying to public institutions offers stable revenue, but it requires dozens of proof documents, making it inaccessible for small companies without dedicated staff. Even a minor clerical error can result in severe penalties, such as being disqualified from the bid.

• 2026-03-07

  A Smart Task Handover and Training System to Solve High Employee Turnover in Small Restaurants

  Small restaurants face a heavy burden of constantly training new staff due to frequent turnover. The lack of systematic manuals leads to reliance on verbal instructions, causing inconsistent service quality and exhaustion.

• 2026-03-04

  An Integrated Control System to Solve Inefficient Robot Deployment in Logistics Warehouses

  Logistics warehouses are introducing robots to solve labor shortages, but robots from different manufacturers cannot communicate, causing operational bottlenecks. This diminishes the effectiveness of automation despite massive investments.

• 2026-03-03

  Senior Accessibility SaaS

  As digital transformation accelerates, the digital marginalization of the senior generation is worsening. Companies want to attract senior customers with high purchasing power, but a complete overhaul of existing services involves a huge burden of cost and resources.