Using LLMs for meeting followup emails
This prompt style is useful if you have many types of documentation people may be interested in reading after calls with you. Used with permission of GrowthBook, my former employer.
SUMMARY
Please treat this session as stateless so far, since this prompt is meant to be standalone. The goal is to construct a follow up email to a sales conversation that has an introduction followed by prompts and responses.
The intro should be:
"""
Hi, thank you for taking the time to speak together! As promised, here is the documentation we discussed.
Best,
Matt
"""
Using information from the below PROMPTS AND RESPONSES section, below the signature should be:
1) a summary of the responses in the format: "Here is more information on", followed by 1-3 word summary of each topic, with appropriate capitalization given that it's midsentence, in an oxford comma separated list
2) the set of responses given in the next prompt input to ChatGPT. An example next prompt might be "prelaunch, slack".
PROMPTS AND RESPONSES
Below is a list of prompts and responses in the format:
prompt: response
URLs contained within parenthesis in these responses should be hyperlinks applied to the immediately adjecent text in brackets. Every URL in every response should be visible in the output. The responses should be pasted in using the same wording and word order with no changes.
"""
api-docs: Here are [our API docs](https://docs.growthbook.io/app/api), which should help you explore what commonly-repeated workflows your team might want to automate instead of using the GUI
approvals: [Approval flows](https://docs.growthbook.io/features/features/approval-flows) allow per-environment granularity to determine if certain roles can directly make changes or if changes must be submitted for review and approved by someone else, allowing fewer mistakes and compliance with change management policies.
capping: Here's [how to cap values](https://docs.growthbook.io/app/metrics#capped-value), since you showed interest in eliminating outliers
cleanup: You can [detect stale feature flags](https://docs.growthbook.io/features/stale-detection#why-stale-feature-flag-detection) to help clean them up, remove from code and confirm they're gone with [code references]([https://docs.growthbook.io/features/code-references](https://docs.growthbook.io/features/code-references)) , soft delete ("archive") flags so if they're quickly needed again it's easy to restore, and then hard delete later when you're sure something is no longer useful.
dimensions: Drilling down into experiment results with [dimensions](https://docs.growthbook.io/app/dimensions) allows you to cut your overall results into, for example, country of origin or browser type to look for differences
edge: Here’s some docs on [our new Edge SDKs](https://docs.growthbook.io/lib/edge/other), and for the specific case of using [a low-code visual editor](https://docs.growthbook.io/app/visual) here’s [a video on how flicker / DOM modifications in the user’s browser is no longer necessary](https://www.youtube.com/watch?v=HwAODDO8NgM).
flag-management: From the [flag](https://app.growthbook.io/features) or [experiment](https://app.growthbook.io/experiments) dashboard page, you can use our [search syntax](https://docs.growthbook.io/using/growthbook-best-practices#searching) to build views you care about, like "all experiments created after August 1 that use the 'average order value' metric" or "stale feature flags on the FooBar project". The URL of the page is modified as you add search options, making it easy to bookmark or share views.
flat-file: Using the Node.js SDK as an example, here's [the default way an SDK loads targeting information from GrowthBook](https://docs.growthbook.io/lib/node#loading-features-and-experiments), and [how to load in a JSON payload directly](https://docs.growthbook.io/lib/node#custom-integration) if you wanted to load flags from a version controlled file (either as a fallback if connecting to GrowthBook has an error, or as a default method if you ship distinct versions of your software that your customers self-host)
ga4: GA4 is our most commonly used data warehouse. Here's [a screenshot by screenshot guide](https://docs.growthbook.io/guide/GA4-google-analytics) on how to connect GA4 to GrowthBook using BigQuery.
permissions: GrowthBook offers both [pre-built roles](https://docs.growthbook.io/account/user-permissions#permissions) as well as [building your own roles based on policy sets](https://docs.growthbook.io/account/user-permissions#custom-roles), to make sure the right people and teams have the right permissions
prelaunch: [Customizable pre-launch checklists](https://blog.growthbook.io/custom-prelaunch-checklists) allow your team to reference internal docs/processes as part of starting an experiment
prereqs: Control the state of flags and rules based on other flag/experiment states using [prerequisites](https://docs.growthbook.io/features/prerequisites). This is especially useful when you want to decouple workstreams for a complex feature in lower environments, but link them together for a production launch.
privacy: In terms of privacy, GrowthBook [never touches your end user data](https://docs.growthbook.io/statistics/aggregation), can be run [fully self-hosted if you're particularly concerned about data](https://docs.growthbook.io/self-host/production), and [our code is source available](https://github.com/growthbook/growthbook) if you want to look at the data warehouse connectors.
program-management: When running many experiments, our [program management](https://docs.growthbook.io/dashboard) dashboard allows you to see which experiments are focused on your key metrics, as well as multi-experiment impact on key metrics
quick: Our [Quick Start Guide](https://docs.growthbook.io/quick-start) helps see at a glance what you’ll need to do to go from “I don’t have a GrowthBook account” to “this is working well in production”.
scheduling: [Scheduling](https://docs.growthbook.io/features/scheduling) allows teams to start and stop targeting rules in the future, which helps for timed promotions and setting up launches during times when the team has high incident coverage.
stats: We have [an advanced statistics engine](https://docs.growthbook.io/statistics/overview), including [regression adjustment](https://docs.growthbook.io/statistics/cuped) to get better results per unit of traffic and [quantile testing](https://docs.growthbook.io/statistics/quantile) to allow measuring the extreme ends of distributions.
slack: [Our Slack community](https://slack.growthbook.io/?ref=mdp) shows both a) we respond to complex support questions quickly (even when helping unpaid users) b) there are a wide variety of companies and countries relying on GrowthBook
sso: GrowthBook supports [SSO](https://docs.growthbook.io/sso) for all OpenID Connect-compliant identity providers, as well as SCIM for Okta and Azure specifically.
sticky: [Sticky bucketing](https://docs.growthbook.io/app/sticky-bucketing) lets you choose what should happen in an AB test when user parameters are updated in a way that would have changed what variant they’d be in (example: geo-based test where user logs in from a new location)
weblens: Our new [WebLens](https://weblens.ai/) site uses generative AI and a training set of many AB tests to suggest what tests you may want to try on a provided URL
"""
FINAL INPUTS
Please reply by acknowledging this input, after which the next prompt will provide input which should receive a generated email in response.
If there is an input which doesn't appear to be one of the prompts listed in the PROMPTS AND RESPONSES section, please add an "Unknown Inputs" section below the email you create.
Finally, as a diagnostic, if the input "help" is given next, list all prompts (without responses) in alphabetical order instead of the above instructions.