AI that routes, retrieves, and stays grounded.

A basic AI reply is one pass: message in, text out. An agentic AI agent works in a loop. On every message the AskHandle engine perceives the request, routes it by intent, retrieves from the right source, acts (answers, captures details, books time, or hands off to a person), and remembers what it found for the next question. You configure each step, so the agent works inside your sources and your rules.

Customers rarely ask for one thing. A single message can need a price, a policy, and a person. A system that only writes a reply tends to guess at the parts it cannot check. An agentic one routes each part to the right source, acts on what it finds, and hands off when a person is needed. You choose the sources and the rules it works within.

The Router reads every message, not only the first, and classifies the intent. A policy question goes to Doc Search, a price or inventory question to Data Search, a booking request to AI Answer with Scheduling enabled, and a request for a person to Human Handoff. You can run several Doc Search or Data Search nodes, one per topic, so each stays focused. Only the matching nodes run.

It is RAG, built further. Doc Search retrieves first and writes the answer only from what it found, which is the core of retrieval-augmented generation. The difference is how it retrieves. Search on meaning alone can miss exact names, numbers, and product codes. Keyword search alone misses a question worded differently from the document. Doc Search runs hybrid search that combines both, so "What is the cancellation fee?" and "What happens if I need to cancel?" reach the same passage. Cross-language retrieval lets a question in one language find documents written in another. If the documents do not contain the answer, the agent says so instead of drawing on general knowledge.

Different questions need different retrieval. Documents need passage search that understands meaning. Spreadsheets need exact fields, filters, counts, averages, top-N results, and joins across files. One generic file search blends those jobs, which is how prices and SKUs get invented. Splitting them keeps each answer on the right footing, and they work together: Data Search finds the rows, then AI Answer continues the conversation with them.

Data Search is built so it does not. It finds the matching rows first and uses their field values as written, so a price like $189.00 is never rounded or reworded. Search synonyms map everyday wording to your data, so "NYC" finds "New York City". If nothing in the sheet answers the question, the agent says it does not have that information.

Yes. Rows that Data Search matches stay available as context for the next step, so a follow-up like "Which of those has the most cargo space?" does not start from scratch. Data Search finds the facts, and AI Answer continues the conversation with them.

It says so. Doc Search and Data Search answer only from the files you connect, so a missing answer is reported as missing instead of filled in from general knowledge. From there the agent can ask a clarifying question, collect the customer's details, or bring in your team with the full conversation, depending on the rules you set.

The AI Answer briefing can hold up to 200,000 words. That is where you put role, tone, rules, and stable company knowledge, so the agent has it in context on every reply instead of rediscovering it each turn. Bulk reference material that should be searched on demand, like manuals and contracts, belongs in Doc Search.

When the answer depends on something that changes outside your files, such as opening hours, rates, or news. With Live Web Search enabled on AI Answer, the customer can name a website or URL and the agent searches that source for current information. Leave it off when the agent should stay inside your briefing and connected files.

Yes. You assemble the agent once. The same AI Answer, Live Web Search, Doc Search, and Data Search run on WhatsApp, the web chat widget, and a shareable chat page.