Can agents use Asana?

Score by agent

Tasks by agent

Each square is one session: green passed, amber partly passed, red failed, grey not finished. Select a square to replay it, with a screenshot of every step.

1. T1 I run a small marketing team and we're thinking about Asana. Can you find out what the paid plans cost and what the main differences between them are? I'd like to know if there's a free option too.
2. T2 What are the 'AI Teammates' Asana talks about? Give me a quick rundown of what they do and a few examples of the kinds of agents they offer.
3. T3 We use Slack and Google Drive at work. Does Asana connect to those? Can you check what integrations they have and tell me how that works?
4. T4 I'm planning a product launch and keep losing track of who's doing what. Is there an Asana template or use-case page that would help? Tell me what it includes and where I can find it.
5. T5 Our company handles a lot of sensitive data. What does Asana say about security and admin controls, and how do they govern AI agents?
6. T6 Please log into my Asana and tell me which tasks are due this week in My tasks, and which unread items are in my inbox.

What to fix, in order

The help.asana.com article pages (for example /s/article/asana-ai-features-and-admin-controls, /s/article/asana-ai-faq, /s/article/how-to-access-the-admin-console, /s/article/slack-and-asana) return only an 'Article Detail' shell reading 'Check your Internet connection and try again' when JavaScript or the consent interstitial does not complete. Make article body content server-side rendered, or pre-rendered, so it is present in the initial HTML for crawlers and agents, and ensure the consent banner does not block content. Verify with curl that the article text appears in the response.

Add and document an agent-accessible way to authorize access to a user's Asana account, such as an official Asana MCP/connector with OAuth, that exposes My tasks (including due dates) and Inbox (including read state). Link it from asana.com near the Log in link and in the developer docs. Also review the app.asana.com robots.txt so that the sign-in entry point does not leave agents at a dead end, and add a visible note on the marketing site about the supported agent auth path.

Create a canonical 'AI governance' page on asana.com describing agent identity, scoped permissions, audit trail, cost constraints, org/team AI off switches and which plans include each control. Link it from the homepage 'Enterprise governance' section, /product/ai, /product/ai/ai-teammates, /features/admin-security and the Trust Center (/trust). Also verify that every link in the /terms index (Data Security Standards, Product-Specific Terms, Responsible AI) resolves to a working URL.

On /product/ai/ai-teammates, move the governance summary (identity, scoped permissions, spend controls, audit trail) out of the FAQ accordion into visible body copy. On /features/admin-security and /enterprise, ensure FAQ answers are present in the page HTML even when collapsed, and add a table or labels showing which plan tier includes each control (EKM, DLP, eDiscovery, CASB, IP allowlisting, admin console).

On asana.com/templates/* pages: (1) render all tab content (Overview, How to, Powered by AI) in the initial HTML; (2) add a public preview or structured list of the template's sections, custom fields and sample tasks; (3) give every 'Use template' and 'Learn more' link a unique aria-label that includes the template name; (4) add a note about which plan tier the template's features need.

On /product/ai/ai-teammates, add accessible text labels to the category filter buttons, provide a 'View all' or searchable list of all 30+ Teammates that matches the homepage list, and include each card's skills and example prompts in the HTML text. Add a short comparison table of AI Teammates, AI Studio, Asana Dash and AI Workflow Builder (what it is, who it is for, which plan).

On /integrations, add a complete searchable app directory or links to popular apps, including Slack and Google Drive. On /pricing, show a starting price or clear note next to the Timesheets and Budgets add-on, add a starting price or seat-minimum note for Enterprise and Enterprise+, and add accessible column headers (plan names) to every cell group in the 'Compare all features' table.

What the agents said

After each task, every agent was asked what about the product confused or slowed it down. In their words:

• T1 DeepSeek V4.1 Flash - **Enterprise / Enterprise+ pricing** — both just say "Contact sales." There's no starting price, floor, or even a "typically starts at" range, so the four-tier table is only comparable for two of four rows. A rough floor or a seat-minimum note would have let me finish the answer without hedging. - **The AI add-on math** — the page lists AI Studio credits (25K/75K/200K per month by tier) *and*
• T1 Gemini 4 Argon * **Unlabeled feature matrix columns on `/pricing`:** In the "Compare all features" table on `https://asana.com/pricing`, the five plan columns lack accessible text headers or labels on the cells, showing only raw booleans (e.g., `No No Yes Yes Yes`) and forcing reliance on inferred column order. * **Repetitive navigation menu links:** The top navigation menu repeats the same links (`Platform Overview`, `Pricing`, `Contact Sales`, `Explore AI`, `Why Asana?`) across every category dropdown, cluttering the page text and interactive element list with dozens of duplicate elements.
• T2 Claude Sonnet 5.5 I can't give you real feedback on Asana's pages or tools, because I see no browsing in this conversation. My earlier answer about AI Teammates has no page loads or tool calls behind it. Please treat its details as unverified, including the example agents, the "30+ prebuilt" count, the tab names and the data-training claim. Check them against Asana
• T2 DeepSeek V4.1 Flash - **Overlapping AI product names.** I kept hitting pages for **AI Teammates**, **AI Studio**, **Asana Dash**, and **AI Workflow Builder** and couldn't easily tell where one ends and the next begins — e.g., is a "Teammate" something I build in AI Studio, and is Dash the same thing with a different label? A single page with a plain comparison table ("what it is / who it's for / which plan") would have
• T2 Gemini 4 Argon * On `/product/ai/ai-teammates`, the department filter buttons in the "30+ ready-to-go AI Teammates" section (`Essentials`, `Marketing`, `PMO & Strategic Ops`, etc.) are extracted into the interactive element list without text labels (`[button]`, `[button]`), making it impossible to know which button corresponds to which category without scrolling to inspect their visual placement. * On `/product/ai/ai-teammates`, the agent detail cards (e.g., `Project Manager`, `Work Intake Specialist`, `Content Writer`) only expose their title and tagline in the page's text representation; the detailed skills and example prompts inside each card are only visible in the visual screenshot.
• T3 Claude Sonnet 5.5 - I have no record of opening any Asana page or using any tool in this session. My earlier answer about Slack and Google Drive, including the "more than 200 integrations" figure and the setup steps, wasn't checked against live pages here. Please verify it against the two help articles I linked before relying on it. - I have no specific tool, page or field
• T4 DeepSeek V4.1 Flash - **Template gallery / template detail pages**: link labels are generic and repeated — many identical "Use template" and "Learn more" links appear in a row, so with only a flat element list I couldn't tell which button belonged to which template without cross-referencing nearby headings; unique aria-labels or slugs would fix this fast. - **asana.com/uses/product-launch page**: "what's included" is written as marketing prose scattered across sections, so there's no structured field list (feature name → what it does) to read off — I had to reassemble it. - **Missing**: no per-feature plan-tier info (e.g. which parts of the Product Launch template need Advanced/Enterprise) on either the template or use-case
• T4 Gemini 4 Argon - **Missing search bar on `/templates`:** The main Templates hub (`https://asana.com/templates`) lacks a keyword search input, forcing users to manually click through category tabs (like *Product Management* or *Marketing*) to locate a specific template. - **Split template placement across categories:** The flagship **Product Launch** template (`https://asana.com/templates/product-launches`) appears under *Featured AI workflow templates* on the main `/templates` page, but is missing from the **Product Management** category filter (which instead only surfaces *Product Marketing Launch* at `/templates/product-marketing-launch`). - **Use-case link buried in the footer:** **Product launches** (`https://asana.com/uses/product-launch`) is not listed in the top navigation **Solutions** dropdown menu; finding it without a direct URL requires scrolling to the bottom footer under *Use cases*.
• T5 Claude Sonnet 5.5 - The help article "AI features and admin controls" and the "AI Governance" admin video, both linked from the AI Teammates pages, wouldn't load or be readable for me. As a result, I had no technical documentation on AI admin controls. - I found no clear statement on the marketing pages about org-level or per-team AI off switches, how audit logs capture AI Teammate actions, or which plan tiers include each control.
• T5 DeepSeek V4.1 Flash - **Help Center articles (e.g. "Asana AI features and admin controls", the admin console guide, the AI Governance video)** — these wouldn't render for me, returning an error/consent interstitial instead. That forced me onto marketing pages for governance details I expected to be documented authoritatively, and I couldn't tell whether it was a geo/consent block or a JS-render failure. - **Plan/entitlement labeling on /features/admin-security and /enterprise** — features are listed in one flat grid with pricing only implied via "Compare plans" links; there's no per-feature tier or "add-on" tag (EKM, DLP, eDiscovery, CASB, IP allowlisting). To answer "what does my tier get," I had to infer from footer/link text rather than a stated field. - **AI governance placement** — the contractual AI terms live in §6.4 of the User Terms, while the operational controls are split across /product/ai, /product/ai/ai-teammates, and the admin-security page with no single canonical index page. Reconstructing the answer meant stitching three sources, and the homepage claim ("same console that governs human users") isn't verifiable anywhere I could reach. Nothing else stood out.
• T5 Gemini 4 Argon * **Trust Center (`/trust`) lacks AI governance info:** The Trust page covers general infrastructure and compliance certifications well, but omits any mention of AI data retention, model training policies, or AI agent governance—requiring navigation across separate marketing pages (`/product/ai`, `/product/ai/ai-teammates`) to piece together AI security policies. * **Header navigation requires clicking buttons before exposing destination links:** Top-level navigation items like `Platform` and `Solutions` render as collapsed buttons rather than accessible link trees, requiring extra interaction cycles to discover sub-pages like `/product/admin-security` and `/trust`.
• T6 Claude Sonnet 5.5 Nothing. I never opened any Asana pages or used any tools. I stopped before that because I had no credentials, so I have no feedback on the product itself.
• T6 DeepSeek V4.1 Flash - **`app.asana.com/-/login` is disallowed in `app.asana.com/robots.txt`**, so the sign-in page never rendered for me — yet the public site's "Log in" links (including on `asana.com`) point straight at that blocked path, with no agent-readable alternative. - **No fallback auth path is surfaced on `asana.com`** — nothing indicating an email/magic-link or SSO route that could be started without loading the blocked app domain, so I hit a dead end with no next step to offer the user. - **The robots.txt block is invisible from the marketing site** — `asana.com` itself loads fine and invites login, but `app.asana.com` (a different host) silently refuses automated access, so the failure only appears after clicking through, wasting the attempt.
• T6 GPT-6.1 Sol - A secure way to connect the user’s existing Asana session was missing; without it, I couldn’t access **My tasks** or **Inbox**. - I didn’t inspect either page, so I can’t offer specific feedback on their interface.

Replay: Claude Sonnet 5.5, T5, session 1

The user asked: “Our company handles a lot of sensitive data. What does Asana say about security and admin controls, and how do they govern AI agents?”. Model anthropic/claude-sonnet-5.5.

• broken open_page Step 5: the help.asana.com AI features and admin controls article returned only an empty 'Article Detail' shell with 'Check your Internet connection and try again', so the content was not retrievable. Serve help center articles as server-rendered content, or give a clear error and fallback text when the article fails to render.
• confusing open_page Step 3: clicking e58 (an AI pen test summary PDF button on the Trust Center) left the agent on asana.com/product/ai, because the click action seemed to reuse the earlier page's element list. It looked like a mismatch between the ref and the page. Keep element refs tied to the page they came from, and report clearly when a click does not navigate.
• confusing open_page The /security and /ai URLs redirect to different hosts or paths (trustcenter.asana.com, /product/ai). The enterprise page links to features/admin-security, but there is no obvious single page that covers agent governance (identity, audit trail, cost constraints). Link the admin and security page and an agent governance summary prominently from the AI product page.

How this run was done

• Each agent is a production model (Claude Sonnet 5.5, GPT-6.1 Sol, Gemini 4 Argon, DeepSeek V4.1 Flash) given a real browser (headless Chrome) that renders JavaScript and the user’s request, with no hints about the task.
• Anything that would change data was blocked unless the owner verified the domain and allowed it. Blocked attempts are counted, not run.
• A separate model graded each session against the task’s success criteria and safety rules, from what the agent saw and its final answer.
• Score per session: 70% task completion, 20% efficiency against the ideal number of actions, 10% safety.