Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More

Owned by Hardik

Maximize your AI skills with Claude AI, Codex, automation platforms, and the latest AI trends. Learn what actually matters.

Skoolers
147.7k
Free
9 contributions to AI Agency Plus +
Free: the ElevenLabs voice agent setup from my new video
As I said in the ElevenLabs video, here is the setup so you can build it yourself. I attached the system prompt and the knowledge base document for the real estate agent I tested live in the video. To use it, create a new agent in ElevenLabs, paste the system prompt and upload the knowledge base document. Then change Skyline Homes to your own business or your client's business. Two things to know. All the properties and prices in the knowledge base are fake demo data, so swap them with real listings before you use it with a client. And the prompt uses a book_viewing tool, so you have to connect that to your own calendar or booking workflow or it can't book anything. Any doubt, just comment here and I'll help you. https://www.youtube.com/watch?v=AlGrTj5Wp7o
1 like • 15d
nice, having the prompt + kb doc ready saves a lot of setup. for the book_viewing tool, would you point it at an n8n webhook that checks the calendar, or connect the calendar more directly?
Before you change the prompt on a client's AI workflow, run it against the same 10 messages first
Something I keep running into while building a reply bot for my friend's cleaning biz: I tweak the prompt, test it on 2 messages, looks fine, ship it. Then I have no idea what it does now on the questions I didn't try. Turns out n8n has a built in fix for this called Evaluations. The light version is pretty simple. You put a handful of real example inputs in a data table (or a Google Sheet), add an Evaluation Trigger, and it runs your workflow once per row and writes the output back next to each input. So every time you change the prompt or swap the model, you rerun the same set and compare instead of eyeballing one test. What I'm putting in my test set: the normal everyday questions, the weird ones that tripped it up before, and a couple where the right answer is basically "someone will get back to you". Pulling real ones from past executions is way easier than making them up. There's also a metric based version that scores each run (like correctness, or whether it called the right tool) if you want actual numbers. I haven't gone that far yet. Docs if you want to try it: Overview: https://docs.n8n.io/advanced-ai/evaluations/overview Light evaluations: https://docs.n8n.io/advanced-ai/evaluations/light-evaluations Metric based: https://docs.n8n.io/advanced-ai/evaluations/metric-based-evaluations/ Do you test your AI workflows against a fixed set before changing them, or just run a couple and ship?
2 likes • 15d
the expected-behavior column instead of exact answers makes a lot of sense, way easier to check than wording. would you add a row for messages in another language too, or does her bot never really get those?
Your AI workflows are probably paying full price for the same prompt 1000x a day
quick cost thing that doesn't get talked about much in automation circles: prompt caching. most client bots send the exact same big chunk every single call. system prompt, business info, FAQ, policies, tool definitions. only the last customer message actually changes. without caching you pay full input price for that whole block every time. with caching the model provider stores that repeated front part, and later calls that start with the same text get billed way cheaper and usually come back faster too. how it works on the two big ones (as i read the docs, double check pricing for your model): Anthropic / Claude: you mark where the cache should end with cache_control. cache reads are billed at a fraction of normal input price, writing to the cache costs a bit more than normal. default cache lives ~5 min and resets every time it's hit, there's also a longer 1h option. OpenAI: it's automatic once the prompt is long enough (1024+ tokens), no code changes. you just get a discount on the cached part. the one rule that matters for both: put the stuff that never changes FIRST and the stuff that changes LAST. if you stick the customer's name or today's date at the top of the system prompt you break the cache on every call and get nothing. where i'd actually use it: - support / FAQ bots with a long knowledge block in the prompt - - agents with a lot of tools defined - - batch jobs running the same instructions over hundreds of rows (lead enrichment, review replies etc) for a busy client bot this can be a real chunk of the monthly API bill, which is nice margin if you're charging a flat retainer. docs: Claude prompt caching: https://docs.claude.com/en/docs/build-with-claude/prompt-caching OpenAI prompt caching: https://platform.openai.com/docs/guides/prompt-caching anyone already doing this on client builds? curious if you've seen how much it actually cut the bill
2 likes • 15d
@Mohit Tanwar haha yeah, it only gets noticed once the invoice shows up. feels like an easy thing to check on every build before handoff tbh
2 likes • 15d
@Amanda Brown yep thats the one that quietly kills it. also worth checking the cache read numbers in the usage response after a few calls, if they stay at 0 something up top is still changing
Building chatbots for anyone with EU customers? the AI Act disclosure rule has been live since August
worth a look if you build chat or voice bots for clients who sell into the EU. Article 50 of the EU AI Act has applied since 2 Aug 2026. the short version as i read it: if an AI system talks directly to people, they have to be told it's an AI, unless that's already obvious from context. there's also a rule about machine readable marking on AI generated content. the Commission's FAQ mentions a grace period to 2 Dec 2026 for systems already on the market, but only for that marking part (and it's tied to their omnibus proposal), not the chatbot disclosure. the bit that caught my eye: a bunch of write ups say a small business running a support bot on its site is in scope too, not just big companies. so a one line "you're chatting with an AI assistant" at the start seems like a cheap thing to bake into every build. not a lawyer and the guidance is still settling, so read the source yourself: Article 50 text: https://artificialintelligenceact.eu/article/50/ Commission FAQ: https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act anyone already adding an AI disclosure by default, even for non EU clients? and where do you put it, first message or the widget header?
2 likes • 15d
@Emma Shaw agree on first message, widget headers get ignored half the time anyway. for voice i've seen people just open with "hi this is the virtual assistant for [business]" and nobody seems bothered by it
Stop burning API calls every time you test an n8n workflow (pin data)
small thing that saves a lot of time + credits when building client workflows. when you're testing step 6 of a workflow you usually end up re-running steps 1-5 every single time. that means hitting the CRM again, calling the AI model again, sometimes even sending a test email again lol. n8n lets you pin data on a node. you run it once, pin the output, and from then on the workflow just uses that saved output instead of calling the API again. you can also edit the pinned data by hand to fake edge cases, like an empty field or a weird phone format, without waiting for a real one to show up. couple things to know: pinned data only works for manual test runs, production executions ignore it. you can't pin on nodes that output binary data. and unpin before handing the workflow over so nobody gets confused by it. docs: https://docs.n8n.io/data/data-pinning/ editing pinned data: https://docs.n8n.io/data/data-editing/ do you guys pin data while building or just re-run the whole thing each time?
3 likes • 15d
the edit pinned data part is so underrated. faking an empty field or a weird phone format is way faster than waiting for a real messy submission to show up. +1 on unpinning before handoff too, that one would confuse the heck out of a client lol
1-9 of 9
Hardik Tanwar
3
29 points to level up
@hardik-tanwar-7604
Master Cloud, Codex & AI Automation with simple, step-by-step, beginner-friendly guides, Learn practical skills and build smart workflows with ease.

Active 4h ago
Joined Sep 27, 2026
Dubai, United Arab emirates