## AI Workflow Assistant — Document understanding wired into an existing business process.

- URL: https://www.naveentehrpariya.dev/projects/ai-workflow-assistant
- Industry: AI / Operations
- Client type: SaaS Company
- Country: United States
- Role: Full Stack Product Engineer
- Duration: 4 months
- Status: Live
- Year: 2025

An assistant that reads incoming operational documents, extracts the fields a team actually keys in by hand, and routes anything it is unsure about to a human, with the source passage attached.

### Problem

An operations team was reading PDFs and emails and retyping the same eight fields into an internal tool. It was slow, it was inconsistent between people, and a generic chatbot pilot had failed because nobody could tell where an answer came from.

### Objectives

- Extract structured fields from unstructured documents
- Make every extraction traceable to its source text
- Route low-confidence results to a human instead of guessing
- Keep inference cost proportional to document volume

### Solution

Documents are chunked and embedded into a vector index; extraction runs as a structured tool call against Claude with a strict schema, so the model returns typed fields rather than prose. Each field carries the passage it came from and a confidence value, and anything below threshold lands in a review queue where a human correction is captured as training signal. Responses are cached by document hash so re-processing costs nothing.

### Features

- Structured field extraction with a typed schema
- Source-passage citation on every extracted value
- Confidence thresholds with human review queue
- Semantic search across the processed corpus
- Prompt and cost caching by document hash
- Per-team usage and spend reporting

### Stack

  - Frontend: Next.js, React, Tailwind CSS
  - Backend: Node.js, Express
  - Database: PostgreSQL, pgvector, Redis
  - Cloud: AWS, Vercel, Docker
  - Integrations: Claude API, OpenAI embeddings, LangChain

### Engineering challenges

### An assistant nobody trusted

Problem: The earlier pilot returned confident prose with no provenance, so reviewers had to re-read the document anyway and the tool saved nothing.

Approach: Every field returns with the exact passage it was drawn from, turning review into a glance at a highlight rather than a full re-read.

### Cost that scaled with retries, not with work

Problem: Re-processing a document after any pipeline change re-paid for inference that had not changed.

Approach: Results are cached against a hash of the document and the prompt version, so only genuinely new work reaches the model.

### Outcome

- Manual re-keying replaced by review of pre-filled, cited fields
- Uncertain extractions surfaced for humans instead of silently guessed
- Repeat processing made effectively free through hash-based caching

### What it taught

The hard part of shipping AI in a business process is not the model. It is provenance, thresholds and a human path for the cases the model should not decide.

Case study: https://www.naveentehrpariya.dev/projects/ai-workflow-assistant
More work: https://www.naveentehrpariya.dev/llms.txt