How AI Agents Turn Specialty Chemical Casework Into an Asset
An agent-based platform matters because it leaves each technical inquiry, trial and field visit behind as a reusable case record. Chat tools that work from one user's pastes and memory leave no company record behind. That holds when the agents work inside the specialty chemical company's own records and each case closes as a document a person approved.
Respondents to McKinsey's 2026 survey most often report scaling AI agents within IT, knowledge management and software engineering (McKinsey, 2026; n=1,719).
Chatbots are the most widely scaled AI tool, with 47% of respondents saying their organizations are scaling them across the enterprise, against about two in ten for AI agents (McKinsey, 2026).
40% of respondents from organizations with annual revenue above 1 billion US dollars report scaling AI agents, up from 27% in the 2025 survey, while smaller organizations stayed flat at 22% (McKinsey, 2026).
Leaders expect agentic AI to have the highest impact in customer support and also see high potential in knowledge management, supply chain management, R&D and cybersecurity (Deloitte, 2026; n=3,235).
Under the US OSHA Hazard Communication Standard, significant new hazard information must be added to a safety data sheet within three months (US OSHA, 29 CFR 1910.1200(g)(5), accessed 2026).
What did McKinsey's August 2026 State of AI survey find about AI agents?
McKinsey's State of AI survey, published on 25 August 2026, found that organizations most often scale AI agents in IT, knowledge management and software engineering. The online survey drew 1,719 participants in 97 nations and was in the field from 4 May to 8 June 2026 (McKinsey, 2026). McKinsey defined agents as "AI systems based on foundation models that act in the real world and are capable of autonomously planning and executing multiple steps in a workflow" (McKinsey, 2026).
The report keeps agents apart from chatbots. Chatbots remain the most widely scaled AI tool: 47% of respondents say their organizations are scaling them across the enterprise, while about two in ten report reaching that stage with AI agents (McKinsey, 2026). Company size also separates the results. Forty percent of respondents from organizations with annual revenue above 1 billion US dollars report scaling AI agents, up from 27% in the 2025 survey, while the share at smaller organizations stayed flat at 22% (McKinsey, 2026).
Deloitte's State of AI in the Enterprise 2026, based on a survey of 3,235 leaders in August and September 2025, points in a similar direction: leaders expect agentic AI to have the highest impact in customer support and also see high potential in knowledge management, supply chain management, R&D and cybersecurity (Deloitte, 2026). Neither survey breaks out specialty chemicals. Read together, they suggest that the handling of a company's own knowledge is one of the first places agent work is landing.
Where does a specialty chemical company's technical knowledge end up?
In many specialty chemical companies, the technical knowledge that daily work produces ends up in email threads, personal files and the memory of the people who did the work. That is an observation about how the work is organized, not a survey result. Picture a technical inquiry about a coating that blisters on a customer's new substrate. The application engineer checks the technical data sheet (TDS), asks about cure temperature and surface preparation, recalls a similar case at another plant and replies. The reply helps the customer. The reasoning behind it, which conditions were checked, which product was ruled out and why, stays in one mailbox.
Trials and field visits follow the same path. A trial report sits in a shared folder under a file name only its author understands. Visit notes stay in a notebook or a phone. The safety data sheet (SDS) revision the answer relied on is not recorded anywhere. When the same customer asks a related question months later, or the engineer moves to another role, the company starts the case over.
The cost is easiest to see when your best technical person leaves, but the loss happens case by case long before that. A customer relationship management (CRM) system usually records that a call took place and a short note, not the reasoning behind the answer, which is why a CRM cannot replace a knowledge base in technical sales.
Why does a chat tool leave no company record for the next specialty chemical case?
A general-purpose chat tool leaves no company record for the next specialty chemical case because each session works from what one user pastes in or what the tool remembers for that user, and ends with an answer copied somewhere else. An engineer can paste a customer's question, two data sheets and a visit note into a chat window and get a reasonable draft in minutes. For one question, that is real help.
The limits arrive with the second and third question. The chat tool does not know which customer asked, what that customer's line conditions are or what was recommended to them before, unless someone pastes it in again or it sits in one user's chat memory, which colleagues cannot open as a company record. The answer leaves the tool as an email, and nothing links it to the account, the trial that followed or the outcome. The work happened, yet the company holds no more knowledge afterwards than it did before. Teams end up pasting customer details into a chat tool every time because the tool has no company record of its own to read.
An agent that works inside the company's records reverses that flow. It reads the account, the project, earlier work logs and the documents on file before it starts, and its output is filed back into the same record. Each case then adds to what the next case can draw on.
How do chat tools and agent platforms compare for specialty chemical casework?
Chat tools and agent platforms differ less in the quality of a single answer than in what remains after it. A specialty chemical technical team handles inquiries, trials and field visits for the same customers over years, so what stays behind matters more than any one reply. The table below compares the two approaches for that kind of casework.
Question | Chat tool used by one person | Agent platform inside company records |
|---|---|---|
Where the context comes from | Whatever the user pastes in, plus that user's chat memory | The account, project, work logs and documents on file |
What remains after the answer | A reply in an email thread | A case document filed with the customer's record |
Which document revision was used | Not recorded | Cited in the case document |
What the next similar case finds | Nothing the team can open, unless someone remembers | Earlier case documents for that customer or application |
Who checks the work | The user, if they notice a problem | A person approves before the document is final |
The row that decides the comparison is what remains after the answer. Both approaches can produce a good reply to one inquiry, and a skilled engineer using either one will often write the same answer. Only the one that files each case back into the company's records turns a year of inquiries into a body of application knowledge that the next engineer can find and reuse. That is also why turning visit and call notes into customer records matters: field visits feed the same record as inquiries, so a trial result and the question that led to it sit side by side.
What turns a specialty chemical case record into a data asset?
A specialty chemical case record becomes a data asset when it holds the conditions, the documents and the reasoning together, so the next similar case can start from it. A record that says only which product was sent is an archive entry. A record that says which substrate and cure schedule the customer runs, which TDS limits were checked, why one candidate was ruled out and what the trial showed is something the next engineer can act on.
Document revisions belong in that record. Safety data sheets change on a regulatory clock. Under the US OSHA Hazard Communication Standard, significant new information about a chemical's hazards must be added to the SDS within three months of the preparer becoming aware of it (US OSHA, 29 CFR 1910.1200(g)(5)). In the EU, REACH Article 31(9) requires suppliers to update the SDS without delay when new hazard information becomes available (European Union, Regulation 1907/2006, Article 31(9)). Both are summaries of the rule text, not legal advice. A case record that cites the revision it used stays checkable after the document changes.
The value compounds with use. The fifth inquiry about the same application can draw on four earlier cases, and an engineer taking over an account inherits its history instead of an inbox. That is the reason every technical case ends in one document: the document is the unit the asset is built from, and each answered case adds one more.
Is a well-kept shared folder enough to keep specialty chemical knowledge?
A well-kept shared folder is enough to keep specialty chemical documents, and a disciplined team with clear file names and a habit of filing trial reports keeps more than most. That deserves credit. A folder costs nothing to start, and every engineer already knows how to use one.
What a folder keeps is files, not cases. It holds the TDS, the trial report and perhaps the visit photos, but the link between them, the customer's conditions and the reasoning that led from those conditions to a recommendation sit in no single file. Filing also depends on someone doing it after the work is done, which is the step that slips when the next inquiry arrives. A folder grows by extra effort. A case record kept by the agent that did the work grows as a by-product of answering.
General-purpose AI tools pointed at a folder help with reading it, but they inherit the same gap: they can summarize the files that exist, not the reasoning nobody wrote down. The folder view is right that structure matters more than software. The difference is whether the structure comes from extra work or from the work itself. A useful line for deciding which tasks to hand to agents is drawn in what to automate in industrial chemistry and what must stay human.
What should an agent platform for specialty chemical casework be able to do?
An agent platform for specialty chemical casework should do the work inside the company's own records and leave each case behind as a document a person approved. Five conditions separate a platform that builds an asset from a tool that answers and forgets:
The agents gather what a case needs from the company's own records first: accounts, projects, work logs and documents such as the TDS, SDS and certificate of analysis (CoA).
Every answer shows its basis, naming the documents and the field history it relied on.
The reasoning from the customer's conditions to the conclusion is visible, and a person approves before the result is final.
Missing information is not made up: an estimate is marked as an estimate with its basis, or the point is left unconfirmed.
Each case ends as one document filed with the customer's record, where the next similar case can find it.
Waiting has a cost that appears on no report. Every inquiry answered only by email is a case that will have to be reasoned again, by someone who may not know it was answered before. The start can be small: one kind of work, such as technical inquiry replies, run through the platform from the first case. The record begins with that case and grows with each one after it.