Use Case · Support

Support deflection on the Software Lifecycle Cloud

Support deflection on KERNSTACK means a first, honest attempt to answer from your own help corpus before a ticket is created: a function retrieves relevant passages from a vector index over your articles, drafts a suggested reply with an inference model, and either resolves the question or opens the ticket with the attempt attached. Questions it could not answer land in a table — that table is the real product.

Functions (WASM)Vector SearchModel InferenceNoSQL Tables
Talk to us Read Technical Documentation → Software Lifecycle Cloud →

The engineering bottleneck with assembled stacks

Support managers do not lack tickets; they lack a filter that respects the customer. A brittle FAQ page deflects nobody, a generic chatbot deflects trust, and agents spend their week answering password-reset-class questions that the help center already answers — while the questions the help center cannot answer are invisible in a sea of closed tickets.

Architecture on the Software Lifecycle Cloud

The widget or help page calls a gateway route; the function embeds the question, retrieves from the article index, and asks a catalogue model to answer strictly from the passages. A confidence path decides: show the answer with its sources, or create the ticket with the retrieved passages attached so the human agent starts from evidence instead of from zero.

Deflection that can be audited

Every deflected answer cites the passages it came from, and every failed deflection is a row in the unanswered table with the question as asked. Weekly, that table tells the documentation team exactly which article to write next — the rare analytics loop that improves both the bot and the humans’ workload at the same time.

  • Answers are grounded in retrieved passages; the function can refuse rather than improvise
  • Escalations carry the retrieval with them, so agents see what the customer was shown
  • The unanswered-questions table ranks documentation work by real demand, not opinion

What to measure after launch

Deflection rate alone is a trap metric — a bot can deflect by being unhelpful. Pair it with reopen rate and with the unanswered table’s growth. On KERNSTACK both come from project tables, so the review is a query over your own records, not a vendor dashboard’s definition of success.

FAQ

Frequently asked questions

A catalogue model for the answer step and text-embed-v1 at 3 credits per 1,000 tokens for retrieval, unless you declare a custom inference endpoint. Model choice is per project and can change without touching the corpus or the widget.

By design, the ticket opens anyway — with the retrieved passages attached for the agent. Deflection is an offer, not a gate: any “talk to a human” click bypasses it instantly, and those clicks are logged as signals too.

Yes. The surface is a gateway route returning an answer and citations as JSON; the widget you already ship calls it like any API. The corpus, index, and logs live behind that one route.

Connected in the Project Graph

Related capabilities, architectures, and comparisons

Category PillarThe Software Lifecycle CloudWhy people and agents build, deploy, and operate on a single vertically integrated stack.Service · ComputeFunctionsGo WebAssembly HTTP handlersService · RetrievalVector searchVector and hybrid retrievalService · ModelsInferenceCatalog and custom model callsService · DataTablesServerless key-value and document tablesIndustry SolutionE-commerce Cloud SolutionsE-commerce on KERNSTACK owns the parts of commerce that constitute the business: the catalogue and its search, carts and checkout state, the order lifecycle, raw payment events reconciled into provable order states, and fulfilment workflows with human steps for the exceptions. Cards stay with the payment provider; the data that makes the business worth buying moves onto infrastructure the merchant controls, in Europe.Use Case · ProductExperiment assignmentAssignment is a pure function of the subject key and the experiment’s salt — stable across requests, no coordination needed. The function records the assignment row once and serves the variant; outcome handlers were already writing events, which now simply carry the experiment id the assignment returned. Analysis reads assignments and outcomes from neighbouring tables with the same keys.Use Case · AgenciesAgency client deliveryThe studio’s delivery shape is a template stack.json: gateway, functions or workers in the client’s stack languages, tables, buckets, and apps under one target per client, on European infrastructure the studio can name in the pitch. Previews run the same declaration; the client’s approval happens in the bound conversation with a live link. Every client’s meter is separate by construction — the studio sees delivery cost per client without allocation spreadsheets.

Ship your stack on the Software Lifecycle Cloud.

Closed preview on dedicated European infrastructure. Bring a repository or start from a blueprint.

Talk to us See pricing & credits →