Finance

On Blits.ai this is an AI agent grounded in a knowledge base that holds the research library, house view and product documents, retrieved with hybrid search (vector plus BM25). Document ingestion covers PDF, Word, PowerPoint and email files, and the document library keeps versions so an outdated note can be reverted or removed. The agent is instructed to answer only from retrieved passages and to cite them.

Advisors reach it in Microsoft Teams or through the API inside their own desktop. Guardrails check inputs and outputs, PII masking keeps client data out of prompts, and role based access controls who can change the agent. Test suites replay real advisor questions with LLM grading against approved answers on every change, and analytics and feedback show adoption and unanswered topics. The platform is model agnostic. Run Blits.ai in our cloud in the EU, Qatar, Singapore, the UAE, the US or Australia, in your private cloud, or fully on premises to meet data residency requirements.

On Blits.ai an agentic workflow runs after period end: custom functions or a SQL knowledge base read performance and holdings from the firm's systems, a knowledge base supplies the house view and manager notes, and an AI agent with structured output drafts each section in the approved template, referencing figures by field. A second function compares every number in the draft with the source data and checks that the required disclaimers are present.

Drafts pause for human in the loop approval, and multi language support and machine translation adapt approved text per market. Guardrails block banned phrases, the workflow's audit trail and downloadable run data serve as the decision log, and test suites replay reference portfolios on every template, prompt or model change.

On Blits.ai the capture step uses speech to text with self hosted transcription and speaker diarization, so audio can be processed on Blits.ai infrastructure. Run Blits.ai in our cloud in the EU, Qatar, Singapore, the UAE, the US or Australia, in your private cloud, or fully on premises. An agentic workflow then drafts the file note in the firm's template with an AI agent using structured output, and pauses for human in the loop approval by the advisor before anything is filed or sent.

Approved notes and tasks are written to the CRM through custom functions (REST calls) or the ready made Microsoft Dynamics 365 tool, and the follow up is drafted for email. PII masking at the gateway and guardrails control what reaches the model, the workflow's audit trail records each run and approval, and test suites check note quality against a set of reference transcripts on every prompt or model change.

On Blits.ai the scoring stays in the firm's own models or data platform; Blits.ai adds the explanation and delivery layer. An agentic workflow runs on a schedule, reads scored signals and client context through custom functions (REST or SQL) or a SQL knowledge base, and an AI agent writes each prompt's rationale and talking points with structured output, grounded in the house view held in a knowledge base.

Prompts reach advisors by email or through the API inside the advisor desktop, and advisors can ask a Microsoft Teams assistant about a client. Agentic tasks can watch for conditions such as a maturity date and fire a prompt when it is reached. Output guardrails with policies the firm writes can block prompts that push products outside policy, and the workflow keeps an audit trail of every run. Act, snooze and dismiss on each prompt is captured and used to improve the ranking.

On Blits.ai the conversation runs in an AI agent that gathers goals and facts, with a flow for the fixed parts of the fact find and custom functions that call the firm's planning engine through its API, so every projection comes from the engine. A knowledge base holds the approved explanations of assumptions and concepts, and structured output keeps numbers in separate fields that a custom function can compare with the engine response.

The advisor version runs in Microsoft Teams or inside internal tools through the REST API channel; a client version can use the web chat channel, or the API channel inside the firm's own app, with rich cards and charts to compare scenarios. Guardrails stop product recommendations and hand complex cases to a human through human handover, and test suites replay known scenarios on every change. Multi language support covers markets where clients plan in their own language.

On Blits.ai this is an agentic workflow that is started through the API with an API token, or from a flow (for example one behind the email channel or the customer facing servicing agent) through a trigger workflow block. The agent reads the request it is given, retrieves the procedure from the knowledge base, and calls custom functions for the entitlement check and the change in core systems, limited by a tool execution policy.

Human in the loop confirmation holds any change above the configured threshold for a checker, and every run keeps a full audit trail. A flow can start the workflow through a trigger workflow block and wait for its result, or confirm completion to the customer by email. Test suites run sample requests against the workflow before a new request type goes live, and PII masking at the gateway can mask or redact personal data the model does not need.