130 practical AI hacks that companies and businesses use or can use, across 11 tools from ChatGPT and Claude to Higgsfield and Grok Bot. Each one says what to do, who it suits and how much effort it takes, and links to the source so you can check it yourself. Last updated 1 November 2026.
Search by tool, tactic or tag, then narrow it down by tool, business area, effort and date.
Tool
Showing 20 of 130 matching hacks (130 in total)
ChatGPT hacksSalesAdded 1 November 2026
Have a team task write the onboarding brief the moment a deal closes
ChatGPT Team Tasks run in the cloud under the team's service account, on a schedule or when an app event happens. OpenAI's own example for sales is to prepare an onboarding brief with customer goals, commitments and open questions when a deal closes. Because the task uses shared workspace connections, nobody has to connect their personal accounts, and a teammate can keep it running when the creator is out.
Effort
Medium
Area
Sales
Best for
Sales and customer success teams on ChatGPT Business or Enterprise
Let Claude flag every clause that breaks your contract playbook
Claude's legal artifact gallery includes a Playbook deviation board. It compares incoming contracts with your company's standard terms, flags each clause that differs and sends it to whoever approves exceptions. Claude does the first pass, a lawyer applies judgment, and every output links back to its source.
Fill a 3,000 cell research sheet and check every cell's source
Fill with Gemini in Google Sheets now shows the exact web sources behind each generated cell, and one request can cover up to 3,000 cells. Use it to enrich a lead list or categorise a product catalog, then spot check the cited sources before anything goes to a client. Cell level citations work in Sheets on the web.
Effort
Low
Area
Research
Best for
Sales ops, marketing and research teams on Workspace
Perplexity hacksCustomer successAdded 1 November 2026
Triage support tickets with probabilities, not paragraphs
Perplexity's Decisions API answers yes or no, multiple choice and rubric questions with a probability for each answer, so routing becomes a simple threshold in your code. Their cookbook asks three questions per ticket and routes each one to escalate, review, queue, auto reply or close. In their sample run only 3 of 12 tickets went on to the pricier Agent API step.
Effort
High
Area
Customer success
Best for
Support teams and SaaS companies with busy inboxes
Voice and translate your ad voiceovers without a recording booth
Higgsfield Audio does text to speech voiceovers, translates audio into other languages and changes voices in one place. Write the script once, generate the voiceover, then make versions for each market you sell in. It saves booking voice talent every time a line changes.
AI image generation hacksMarketingAdded 1 November 2026
Lock your brand look with one reference image and reuse it everywhere
Recraft builds a custom style from one to ten of your reference images, with no training, and captures the rendering, colour, texture and composition. Create it once and pass the same style_id in every generation request so campaign images all look like they came from one designer. Styles from the web app can be shared to other accounts too.
Effort
Low
Area
Marketing
Best for
Design and marketing teams producing lots of on-brand visuals
AI video generation hacksMarketingAdded 1 November 2026
Reuse your best performing ad for every product in the catalog
Runway's Product Swap recipe takes a reference video, a photo of the product in it and one to ten images of a new product. It swaps in the new product while keeping the original camera motion, lighting and scene. Take the ad that already converts and roll it out across your catalog instead of shooting each one.
Write a REVIEW.md so Code Review stops drowning you in nits
Commit a REVIEW.md at the repo root and Claude Code Review follows it on every pull request. You can redefine what counts as Important, cap nits at five, skip generated files and lockfiles, and add must check rules like tenant scoped database queries or no emails in logs. Reviews get shorter and focus on what can actually hurt the business.
OpenAI Codex hacksEngineeringAdded 1 November 2026
Auto assign new Linear issues to Codex with a triage rule
Install Codex for Linear and you can assign issues to Codex like a teammate, or mention @Codex in a comment. Go further with a Linear triage rule that delegates new issues to Codex automatically. Codex posts progress on the issue and finishes with a summary and a link so you can open a pull request.
Effort
Medium
Area
Engineering
Best for
Engineering teams that run their backlog in Linear
Test new software on free trials without the surprise bill
Every team signs up for a trial to test a tool, then forgets and gets billed. Grok Bot's Try it, don't pay for it template starts the free trial for you and then cancels it, or reminds you, before it bills. You get the full evaluation and only pay for the tools you actually decide to keep.
Show ChatGPT a boring task once and keep it as a skill
In the ChatGPT desktop app on a Mac, pick Record a skill and just do the task while it watches, like filing an expense or pulling a recurring report. When you stop, it drafts a skill with the inputs, steps and how to check the result. Next time you only give it what changed, like the date range. Keep secrets and sensitive data out of the recording.
Make Claude show the exact line it got the answer from
Turn on citations when you send PDFs or text documents through the Claude API, and every claim comes back pointing at the passage that backs it, with page ranges for PDFs. That makes contract and policy answers easy for a human to check. Bonus: the quoted cited_text does not count toward your output tokens, so it can cost less than asking Claude to quote things itself.
Effort
Medium
Area
Research
Best for
Legal, compliance and support teams building internal Q&A
Google Gemini hacksEngineeringAdded 31 October 2026
Put the shared stuff first so Gemini caches it for free
Implicit caching is on by default in the Gemini API, and Google passes on the savings automatically when a request hits the cache. To hit it more often, put the big, repeated part of your prompt, like your product catalog or policy doc, at the start and the changing question at the end. Send similar requests close together in time.
Get research back as clean JSON rows, not paragraphs
Add a response_format with a JSON schema to your Perplexity Agent API request and the answer comes back matching that shape. Ask for company name, pricing page and key fact as fields and you can drop the results straight into a spreadsheet or CRM. No more copy and paste out of a wall of text.
Build your own AI brand ambassador and swap the product
Higgsfield's AI Influencer lets you create a character, pick a trending motion or upload a 3 to 30 second clip, and generate a ready-to-post video. You can swap outfits, products or locations while keeping the motion and camera the same. One winning video can become a version for every product in the range. Be upfront with your audience that the presenter is AI.
AI image generation hacksMarketingAdded 31 October 2026
Ask for transparent PNGs and draft on low quality
With OpenAI's image API, set background to transparent and output format to PNG or WebP, and you get stickers, icons and product cutouts that drop straight onto any page or ad. Run your first rounds on low quality for quick, cheap drafts. Only compare higher quality settings for the final picks.
AI video generation hacksEngineeringAdded 31 October 2026
Move off Veo 3.1 Preview before your video pipeline breaks
Google shut down the Veo 3.1 Preview models in the Gemini API on October 22, 2026. If your ad or social pipeline still calls them, switch to Gemini Omni Flash or the GA models on Google's enterprise agent platform. Omni Flash also lets you refine a video over several turns, like replacing an element, instead of regenerating from scratch.
Auto-run your linter every time Claude edits a file
Add a PostToolUse hook with the matcher Edit|Write that runs your lint or format script. Now every file Claude Code touches gets checked right away instead of at review time. Commit the settings so the whole team gets the same guardrail.
Write a rules file so Codex can never run your scary commands
Drop a .rules file in ~/.codex/rules and use prefix_rule to set commands to allow, prompt or forbidden. When several rules match, Codex goes with the strictest one. Admins can enforce restrictive rules for the whole team through requirements.toml. Rules are still experimental, so test them with the built-in match examples.
Let Grok run Python for the numbers instead of guessing them
Turn on the code execution tool and Grok writes and runs Python for precise calculations and data analysis. Give it your CSV columns and what you want, like monthly profit margins, and it computes them rather than estimating. The docs suggest a low temperature for math.
Every hack is written up from the linked source, checked on the day it was added. AI tools change fast, so features, plans and prices can move after that date. Numbers are as reported by the source, often by the vendor itself, so treat them as claims, not audited results. Effort and business area are our own rough judgement. Test before you roll anything out, keep a human reviewing what the AI produces, and follow each tool's terms and your own data rules. Company and product names belong to their owners, and we are not affiliated with any of them.
Cookies and your choices
Visits are counted by Cloudflare Web Analytics, which sets no cookies and stores nothing on your device. An affiliate tracking script from Impact.com also loads on our pages and can set its own cookies so partner companies can credit us for sign-ups. Essential storage is always on so pages load and work correctly. You decide whether your browsing preferences are saved on your device. Read the Privacy Policy.