Stromy
Book a Call
Technology

The Output Layer

From AI analysis to board-ready deliverables.

AI can reason and analyse. But the result still needs to land in the format people actually use — a PowerPoint for the briefing, an Excel for the model, a Word document for the review. Automated document generation closes that gap without manual formatting.

The Formatting Bottleneck

Intelligence that sits in a text box isn't delivered.

Enterprise workflows don't end with a text output. They end with a document that goes to a client, a presentation that gets reviewed in a board meeting, a workbook that feeds into a financial model. Every AI analysis that doesn't reach one of these formats gets reformatted manually — or it doesn't get used.

Python's document libraries solve this. AI generates the content as structured data; Python produces the professionally formatted artefact — applying brand specifications, embedding charts, populating templates, and producing files that look like they were made by a senior designer, not exported from a chatbot.

AI writes, Python formats

The language model produces structured content — analysis, narratives, data summaries. Python handles the professional formatting: slide layouts, table structures, brand styles, document sections. The two roles never mix.

Brand enforcement, not brand requests

Colour palettes, typography, logo placement, and template structure are defined once in the brand specification. Every document produced by every workflow for a given partner applies those specifications automatically — not approximately.

Consistent outputs from consistent inputs

The same workflow run on the same source data produces the same document structure. No formatting drift, no inconsistent slide layouts, no missing sections. Reviewability at scale.

Multiple formats, one workflow pass

A single analysis workflow can emit PDF for the briefing, XLSX for the data model, PPTX for the presentation, and DOCX for the working document — all from the same structured output, formatted appropriately for each context.

Output Formats

Seven formats. One workflow pass.

Each output type is produced by a specific Python library. Each library handles one format's full fidelity — formulas, animations, styles, navigation — not a flat export.

PDF

WeasyPrint / ReportLab

Narrative reports and briefings — the primary delivery format for intelligence products.

  • Executive intelligence briefs
  • Stakeholder analysis reports
  • Strategy audit packages
  • Actor and dossier profiles

PPTX

python-pptx

Chart and visualisation embedding from analysis scripts. Brand-enforced layouts and master slides.

  • Board presentations
  • Stakeholder briefing decks
  • Workshop preparation packages
  • Competitive intelligence slides

DOCX

python-docx

Template-driven generation with dynamic section replacement. Styles and formatting preserved exactly.

  • Interview and workshop packs
  • Working documents and contracts
  • Policy analysis documents
  • Proposal drafts

XLSX

openpyxl

Formula preservation, pivot tables, conditional formatting. Not flat exports — live workbooks.

  • Stakeholder × driver matrices
  • Financial and commercial trackers
  • SOV and media analytics workbooks
  • Data models and forecasts

HTML

Astro / Jinja2

Web-published reports with navigation, filtering, and responsive layout.

  • Interactive intelligence dashboards
  • Published briefing pages
  • Legislative pipeline trackers
  • Internal knowledge bases

Video

Remotion

React-based programmatic video. Git-versioned, CI/CD rendered, brand-compliant animation.

  • Data story animations
  • Workflow explainers
  • Personalised client summaries
  • Branded content production

Video as Code

Remotion: production video from a codebase.

Remotion applies software engineering practices to video production. Scenes are written as React components — composable, testable, version-controlled. An AI agent can generate a fully animated data story or branded explainer by writing components, not editing a timeline.

This changes the economics of video production. A personalised intelligence summary video for fifty stakeholders costs the same as one. Data visualisations that update weekly render from a CI/CD pipeline, not a video editor's workstation. The brand consistency of every frame is guaranteed by code, not by convention.

Scenes and animations are written as React components — versioned in Git, reviewed like code, tested in CI

Data visualisations update automatically from live API data — no manual re-exporting

Voice and transcript timing can be synchronised using AI-generated narration

Renders at scale on cloud infrastructure — no local GPU or editing suite required

Previous layer

Skills & AI Workspaces

How domain expertise is encoded and deployed as a complete AI toolkit.

Explore

Deployment context

Enterprise Integration

How the full system deploys into your M365 environment, ERP, and cloud infrastructure.

Explore

Overview

Technology Platform

See how document automation fits within the full technology platform.

Back to technology

Ready to start

Let's reimagine how your business works.

Whether you need a strategic roadmap or a full implementation partner, we'll help you unlock the potential of AI for your specific domain.

Book a discovery call Send a message