{"id":107928,"title":"Scispot Smart Actions","tagline":"We stopped connecting instruments and started letting them make decisions","body":"![uploaded image](/media/?type=post\u0026id=107928\u0026key=user_uploads/697526/3baf4e61-60db-413e-a024-a596c1cf70d1)\n\nAt 11:40pm, an analyst at a biotech we work with was staring at a plate of 96 ELISpot wells, trying to figure out why three \"positive\" results didn't match the raw counts. She'd already re-typed the spot counts into a spreadsheet once. She was about to do it again, by hand, to double-check herself because the instrument doesn't know what a \"positive\" is. It just counts spots. Someone has to turn that count into a decision, and that someone was her, at midnight, for the fourth night that week.\n\nThat's not a story about one analyst. That's the entire industry's relationship with lab data, and it hasn't meaningfully changed in a decade.\n\n## **The part everyone stops short of**\n\nMost labs eventually solve the easy problem: getting a file off an instrument and into some kind of system of record. That's plumbing, and plumbing gets built. What almost nobody automates is the step after deciding whether a result is even valid.\n\nDoes it clear the CV threshold?\n\nIs this well too confluent to trust (TNTC)?\n\nWhat's the right baseline to compare against? Is this a Positive, a Negative, or a QC Fail?\n\nThat judgment still lives in someone's head and someone's spreadsheet, redone by hand, every single run.\n\nWe think that's the real bottleneck in lab data - not moving files around, but making trustworthy calls on what's inside them, fast enough to keep up with how fast science actually moves.\n\n## **The infrastructure that makes this possible: GLUE**\n\nNone of this works without a real connectivity layer underneath it, so we built one. [Scispot GLUE](https://www.scispot.com/glue) is our API-first integration backbone with hundreds of connectors, native HL7/ASTM support, and webhooks — that plugs into instruments, LIMS-like entities, and external systems without custom code for every new device.\n\nGLUE is the nervous system; Smart Actions are the brain sitting on top of it, deciding what every signal actually means.\n\nThat distinction matters. A lot of \"lab automation\" stops at GLUE's job: get the file, map the fields, store the record. We built Smart Actions specifically because moving the data was never the finish line — it was the prerequisite.\n\n## **What we actually built**\n\n[Smart Actions](https://www.scispot.com/smart-actions) are AI-agent workflows inside Scispot that don't stop at moving data - they interpret it.\n\nA Smart Action parses a raw instrument export, maps every well to the right sample and treatment condition, applies your lab's QC rules automatically (CV thresholds, LLOQ cutoffs, TNTC handling), computes the real statistics — averages, fold change against baseline — and writes the _decision_, not just the number, straight into your Labsheets.\n\nThe ELISpot plate our analyst stayed up for?\n\nToday that run finishes, and by the time she opens her laptop, the QC-gated summary is already sitting in the [Scispot Labsheet](https://www.scispot.com/labsheets) - Positive, Negative, Indeterminate, or QC Fail, each one traceable back to exactly why. Not a spreadsheet she has to trust blindly. A decision she can audit in one click.\n\n## **Why this is a bigger bet than \"faster data transport\"**\n\nMoving data fast, which is what GLUE gives you, is table stakes. The hard, valuable problem is the judgment layer on top of it is this result real, does it pass QC, what does it mean for the next experiment. That's what Smart Actions own, permanently, for every run, not just the one someone remembered to check carefully.\n\nSmart Actions are composable, build one for ELISpot, reuse the same pattern for a plate-reader screen routing hits into follow-up assays, chain them together across protocols, all running on top of GLUE's connectivity. And they're callable by [Scibot Omega](https://www.scispot.com/product-releases/scibot-omega), our AI orchestrator, so a scientist can just say \"clean this run, QC it, summarize it for the sponsor\" and the right Smart Actions execute end to end.\n\n## **Why we're telling you this now**\n\nWe've launched Scispot on YC before as a platform, and GLUE as our connectivity layer.\n\nThis time we wanted to talk about the layer built on top of it that our customers actually thank us for - the layer that means their best analyst stops being a human QC script running at midnight, and starts being a scientist again.\n\nIf you've ever stared at a spreadsheet at 11pm trying to figure out why a result doesn't add up, we built this for you.\n\n👉 **scispot.com/demo**\n\n\u003chttps://www.youtube.com/watch?v=_BFoFo_0zHk\u003e","slug":"S4m-scispot-smart-actions","created_at":"2026-07-27T23:39:25.957Z","updated_at":"2026-09-18T20:15:22.093Z","total_vote_count":14,"url":"https://www.ycombinator.com/launches/S4m-scispot-smart-actions","share_image_url":"https://www.ycombinator.com/media/?type=post\u0026id=107928\u0026key=user_uploads/697526/3baf4e61-60db-413e-a024-a596c1cf70d1","company":{"id":24367,"name":"Scispot","slug":"scispot-io","url":"https://scispot.io","logo":"https://bookface-images.s3.amazonaws.com/small_logos/996a065b9b830e170e3a68e207d6245f28dc54b8.png","batch":"Summer 2021","industry":"B2B","tags":["Generative AI","SaaS","B2B","AI"],"search_path":"https://bookface.ycombinator.com/company/24367"}}