Connect Koji to Make (Make.com): Build No-Code Customer Research Automations

Answer first: Make (formerly Integromat) is a visual automation canvas, and Koji plugs into it cleanly without a dedicated app. Two building blocks do everything: an HTTP module that calls Koji's headless API to start interviews, and a scheduled HTTP module that polls Koji's read endpoint for newly analyzed interviews and routes the insights anywhere: Slack, Notion, a CRM, a spreadsheet, or an email. Koji has no outbound webhooks yet, so the pipeline pulls rather than pushes. That lets you assemble end-to-end research pipelines by dragging modules: when a Stripe subscription cancels, start a Koji churn interview; when it is analyzed, post the themes to Slack and log the row in Notion. It is the same pattern as Koji's Zapier and n8n integrations, with Make's routers and filters for more complex branching. Most scenarios take 20 to 40 minutes to build.

Why use Make with Koji

Make sits between hundreds of apps and lets you wire logic visually — triggers, routers, filters, iterators — without writing a service. For customer research, that means you can put a Koji AI interview at the exact moment a meaningful event happens anywhere in your stack, and then send the resulting insight to the exact place your team already looks. You get the depth of a moderated interview (Koji's AI asks adaptive follow-ups in voice or text and analyzes every conversation automatically) with the reach of an automation platform that already speaks to your billing, CRM, forms, and messaging tools.

Compared to hand-built middleware, Make gives you retries, error handling, and a visual audit trail for free — and compared to a raw survey tool, platforms like Koji turn each triggered touch into a real conversation rather than a form nobody finishes.

The two building blocks

1. HTTP module → start an interview (Make to Koji)

Add an HTTP > Make a request module pointed at Koji's Start Interview endpoint:

• Method POST, with your Koji API token in the Authorization header (see API authentication).
• Body includes the study_id, respondent_email (or respondent_id), and any personalization fields mapped from the trigger. Koji returns an interview link you send yourself (Koji does not send email).

2. Scheduled HTTP module → poll for the insights (Koji to Make)

Add a Schedule trigger (every 15 minutes works for most teams) followed by an HTTP > Make a request module that GETs Koji's read endpoint for the study. Filter to interviews analyzed since the last run, and every module after it can route that data (top themes, sentiment, quality score, structured answers, transcript URL) anywhere. If the interview is embedded in your own product, the embed's completion event can kick off the same scenario.

Three scenario recipes

Recipe A — Churn interview from a cancellation. Stripe (Watch cancellations) -> HTTP (start Koji interview). Then a second scenario: Schedule + HTTP (poll Koji for analyzed interviews) -> Router -> Slack (post themes to #retention) + Notion (append to churn database). Pair with cancel-flow exit interviews.

Recipe B — Post-onboarding interview from your CRM. HubSpot/Salesforce (contact reaches "onboarded" stage) -> HTTP (start Koji interview) -> on completion, Router writes sentiment and themes back to the CRM contact and alerts the CSM. See CRM research integration.

Recipe C — Scheduled batch from a spreadsheet. Schedule (every Monday) -> Google Sheets (iterate new rows) -> HTTP (start interview per row) -> results append to a Sheet via the polling scenario. See Google Sheets + Koji.

Use routers and filters for real logic

Make's routers let one polled Koji interview fan out to multiple destinations, and filters let you act only on the interviews that matter. Common patterns:

• Filter sentiment = negative AND quality_score >= 0.7 -> alert the account owner immediately.
• Route by single_choice root-cause value to different Slack channels or Jira projects.
• Iterate over multiple_choice or ranking answers to write one row per selected option.

This is where structured questions pay off. Each Koji study can mix six question types — open_ended, scale, single_choice, multiple_choice, ranking, and yes_no — and because each returns a typed value, your Make filters and routers can branch on them deterministically. Design them with the structured questions guide.

Reliability and security notes

• Store your Koji API token in a Make connection/keychain, not in plain module fields.
• Keep the last-seen interview timestamp in a Make data store so each interview is routed once.
• Turn on Make's error handlers so a transient failure retries instead of dropping an interview.
• For sensitive studies, run Koji in anonymous mode and route only aggregate insights, never raw transcripts, downstream.

Plans and limits

Koji interviews start as low as €1 per qualified interview, and €3 per qualified voice interview. Start with pay as you go. No subscription needed. You pay only for the interviews your study actually uses. The headless API that Make relies on is included on every plan, including a free account; outbound webhooks are not available yet.

A closer look at the interview payload

The HTTP module returns one JSON object per analyzed interview. Parse it once with Make's JSON module and every field becomes mappable: respondent_email or respondent_id for identity, sentiment and themes for the headline, structured_answers for typed values, quality_score to filter low-effort responses, and transcript_url to link back. From there Make's mapping panel wires each field into whatever downstream module you choose.

When to choose Make vs Zapier vs n8n

All three run the same Koji pattern (HTTP out to start interviews, a scheduled poll to collect insights), so pick by fit:

• Zapier is the fastest to a first working zap and has the widest app catalog; ideal for simple one-trigger, one-action flows. See Connect Koji to Zapier.
• Make shines when a single interview needs to fan out with routers, filters, and iterators, or when you want a visual audit trail and cheaper high-volume operations.
• n8n is the choice when you need self-hosting or want the pipeline in your own infrastructure. See Connect Koji to n8n.

Common pitfalls (and how to avoid them)

• Double processing. Store the last-seen interview timestamp so a poll never routes the same interview twice.
• No error handling. Wrap the HTTP module with a Make error handler and retries so a transient failure does not silently drop an interview.
• Over-broad triggers. Filter to the events that deserve an interview; do not launch one on every record change.
• Token in plain text. Keep the Koji API token in a Make connection, not a bare module field.

Put the interview at the moment that matters

The reason to run Koji through Make is timing. A survey sent days later gets stale, low-candor answers; an AI interview triggered the instant a customer cancels, upgrades, or finishes onboarding catches them while the experience is vivid. Make is the switchboard that makes that timing automatic across your entire stack — and Koji is the AI moderator that turns the moment into a real, analyzed conversation instead of a form nobody finishes.

Build Recipe A step by step

To make the pattern concrete, here is the cancellation-to-Slack scenario end to end. First scenario (trigger the interview): add a Stripe > Watch Events module filtered to customer.subscription.deleted; add an HTTP > Make a request module that POSTs to Koji's Start Interview endpoint, mapping the Stripe customer email into respondent_email and your churn study_id into the body. Save and turn it on — every cancellation now launches an interview.

Second scenario (handle the result): add a Schedule trigger and an HTTP > Make a request module that GETs Koji's read endpoint for the churn study; run one interview so Make can determine the data structure. Add a Router with two routes — route one has a Slack > Create a Message module posting the sentiment and top themes to #retention; route two has a Notion > Create a Database Item module logging the respondent, themes, quality score, and transcript URL. Add a filter on route one for sentiment = negative so only the interviews worth a human's attention ping the channel. That is a complete, production-ready churn-listening loop built from five modules and zero lines of code — and it keeps running whether you are watching or not.

Related Resources

• Connect Koji to Zapier
• Connect Koji to n8n
• Research Automation with Webhooks
• Structured Questions in AI Interviews
• Starting Interviews via API
• Google Sheets + Koji