AI for every scientist. Breakthroughs for all.

Benchling AI brings intelligent agents and cutting-edge scientific models directly into R&D workflows. Rather than forcing scientists to move data into separate AI tools, Benchling AI embeds intelligence where the work already happens. It is available to all Benchling customers, with free access for academic scientists. Learn more about the different features here.

AI Agents are advanced capabilities that retrieve and capture data autonomously (Deep Research, Compose, Data Entry, Ask). AI features are lighter-weight tools that improve everyday productivity (Notebook Check, SQL Writer). Scientific Models are domain-specific predictive tools (AlphaFold 2, Chai-1, Boltz-2, PipeBio protein models). All three are managed by admins and permissions can be enabled for specific users or teams.

Benchling AI currently includes four core agents: the Deep Research Agent (answers complex scientific questions by reasoning over your Benchling data and public literature), the Compose Agent (turns protocols, notes, and attachments into structured notebook entries), the Data Entry Agent (converts unstructured PDFs, spreadsheets, and CRO reports into clean, structured Benchling data), and the Ask Agent (a conversational interface for finding experiments, samples, and results in seconds). See the AI tools for the modern lab blog for a deeper walkthrough of each.

Benchling AI gives scientists direct access to leading structure prediction and generative models including AlphaFold 2, Chai-1, and Boltz-2, without leaving the platform. Model predictions are connected directly to your experimental data in Benchling, unifying in silico design and in vitro work. Benchling is also integrating NVIDIA NIM microservices including NVIDIA BioNeMo models, and supports models from Anthropic Claude, Google Gemini, OpenAI, and Lilly TuneLab.

The Deep Research Agent analyzes internal experimental data and public literature to answer complex scientific questions, producing detailed, citation-backed reports in minutes rather than days. Scientists can ask questions like "trace the lineage of protein X across all experiments," "compare failed and successful transfections in Project Y," or "summarize in vivo studies ST042 and ST043." It supports multi-turn chat, file attachments, and can connect to external systems via Model Context Protocol (MCP). See our most recent blog for more real-world examples and use cases.

The Compose Agent turns scattered bench notes and photos into clean, structured Benchling notebook entries and templates through a simple conversational interface. It's particularly powerful for teams migrating legacy data: one biotech used Compose to convert 20,000 legacy ELN entries into structured Benchling notebooks, making years of experiments searchable and queryable without months of manual cleanup. See our most recent blog for more real-world examples and use cases.

The Data Entry Agent converts unstructured data, including PDFs, Word documents, spreadsheets, and CRO reports, into clean, structured information inside Benchling. Scientists simply upload a file and the agent orchestrates a sequence of LLM calls to plan, process, and verify the output. This is especially useful for teams receiving inconsistent data deliverables from external partners.

Experiment Optimization uses classical machine learning and Bayesian optimization to analyze past experimental results and recommend the next best conditions to test. It helps scientists visualize multi-dimensional experimental spaces, compare conditions, and accelerate decision-making, without needing a dedicated data science team. It's particularly valuable in Bioprocess for process development cycles, and in Bioresearch for assay development.

Benchling is built on the principle that your data stays yours. Benchling does not train its own models on customer data. All AI actions are logged in product audit trails, and AI capabilities are developed under the same secure software development lifecycle as all Benchling features, including SOC 2 Type 2 and ISO 27001 compliance. Full details are available in the AI data protection and security policy and the Benchling Trust Center.

Yes. Benchling AI is designed with open standards in mind. It supports Model Context Protocol (MCP) for connecting to external systems like Microsoft Teams, SharePoint, Slack, and other platforms without custom APIs. It also supports frontier models from NVIDIA BioNeMo, Anthropic Claude, Google Gemini, OpenAI, and Lilly TuneLab directly within Benchling, with no vendor lock-in.

Some Benchling AI features, including Notebook Check and SQL Writer, are included with your Benchling subscription at no additional cost. Agents and models use a credit system, and credits are included with every Benchling subscription so teams can start using AI right away. See the Getting started with Benchling AI help article for current credit information and how to enable AI features from the admin console.

Yes. Benchling has launched benchling.ai as a public entry point to explore Benchling AI, with example queries and interactive previews. Academic scientists also get free access as part of Benchling's academic plan.

The most common reason AI pilots stall in biotech isn't performance — it's data quality. When experimental data lives in spreadsheets, disconnected ELNs, or inconsistent formats, AI can't reliably reason over it. According to Benchling's 2026 Biotech AI Report, the AI use cases with the highest adoption succeed precisely because the underlying data is clean, structured, and verifiable. Benchling AI is built on this principle: agents and models connect directly to structured R&D data in Benchling, so the data problem is solved before the AI problem begins.

AI is only as good as the data it runs on. For biotech teams, that means working on a platform where experimental data is linked to registered samples, annotated with biological context, and captured consistently across teams. Benchling is designed around this from the ground up — every notebook entry, registered entity, and instrument result is connected in a single data model that AI agents can query in real time. See our AI Data Maturity guide for how leading biotech teams are building AI-ready data infrastructure.

AI agents for scientists are purpose-built systems that can autonomously retrieve, reason over, and act on scientific data — not just answer general questions. Unlike general AI tools, they understand biological context: the relationships between experiments, samples, sequences, and results. Benchling AI's four agents — Deep Research, Compose, Data Entry, and Ask — operate directly inside the platform where R&D data lives, so scientists never need to export or reformat data to get an answer. See the AI tools for the modern lab blog for examples of each agent in action.