AI Agents Automation: From Search Prompts to Full Workflows
AI agents automation in Perplexity now combines model-written research workflows with direct design output, letting agents move from gathering information to creating editable assets without manual copy-paste steps between tools. Instead of issuing single search prompts, creative teams can ask for full projects—such as a campaign concept or performance review deck—and let the system plan the research steps, collate findings, and prepare content for design tools. This approach shifts AI from being a suggestion engine into a workflow engine. For creative and marketing teams, the promise is not only speed but also consistency: every project can follow a repeatable pipeline for discovery, synthesis, and asset creation. The open question is how much oversight teams will still need to maintain quality and brand alignment as autonomous agents take on more of the work behind the scenes.
Perplexity Search as Code: Python Pipelines for Deep Research
Perplexity Search as Code lets AI agents write Python search workflows that structure complex research as code instead of repeated API calls. The system relies on a model as a control plane, a restricted compute sandbox, and an Agentic Search SDK that exposes retrieval, filtering, deduplication, and reranking steps. Agents can move beyond a single answer response and instead assemble multi-step pipelines that gather candidate pages, remove duplicates, and rerank evidence before drafting an output. Generated scripts also make the retrieval path visible so teams can audit which sources were considered and how they were ranked. Perplexity reports that in a CVE vendor-advisory task, Search as Code reached 100 percent accuracy while using 85.1 percent fewer tokens than its baseline, though those benchmark claims still need independent validation before teams treat them as reliable for production workloads.
Canva AI Integration: Turning Research into Editable Design Assets
Perplexity’s new Canva connector ties AI-driven research directly to automated design generation. Through Perplexity Computer, users can connect their Canva accounts so AI agents can turn research, analysis, and strategy documents into editable presentations, social media campaigns, infographics, and brand assets. Perplexity Computer can analyse meeting notes, performance data, and web information to create a structured creative brief. That brief is then passed to Canva, where it becomes live design files that remain fully editable. With Brand Kit integration, organisations can apply existing brand elements so AI-produced layouts match their visual identity. According to Canva, the connector lets teams “generate a range of assets including presentation decks, pitch materials, social media content, campaign assets, infographics, data visualisations, brand kits and reusable templates,” cutting out manual transfer from AI outputs into design tools and reducing friction between strategy and execution.
End-to-End Creative Pipelines Without Manual Handoffs
Together, Perplexity Search as Code and the Canva connector enable agents to handle end-to-end creative workflows that span research, synthesis, and asset creation. A typical pipeline could start with an agent running a Search as Code workflow to scan the web, performance dashboards, and internal notes, then organising the results into a campaign narrative or performance story. That structured content passes directly into Canva, where templates for decks, social posts, or infographics are generated and aligned to the brand’s visual system. Because the resulting files are editable, designers can refine layouts and copy instead of starting from a blank page or reformatting AI text. This setup turns AI agents into production teammates for marketing, product, and communications teams that already rely on recurring formats like quarterly reviews, launch kits, and social series.
Cost, Trust, and What Creative Teams Should Watch Next
Perplexity positions its agentic search stack as both more efficient and more transparent than traditional search APIs, with substantial token savings claimed for complex tasks. If repeatable, that efficiency could lower the cost of AI agents automation for creative teams that run frequent research-heavy projects. However, all benchmark numbers so far come from company-run tests, so teams should compare results against alternatives such as OpenAI, Exa, Parallel, TinyFish, and Tavily before standardising any workflow. On the design side, the Canva AI integration promises to reduce manual work but does not remove the need for review: generated search code and design outputs both become part of the trust boundary. Creative leaders will need processes for checking sources, validating key claims, and auditing templates, treating agents as powerful assistants rather than fully autonomous decision-makers.
