I'm Alex Lepinay,
founder of The AI Adoption.

The assistant reviews incoming messages, groups them by urgency and action, and prepares replies using the person’s writing style and work context. The user can work through decisions and drafts together instead of repeatedly rereading the inbox.

Each reply is approved individually. This assistant leaves the inbox undisturbed: it does not send messages, delete emails or follow links on its own. That keeps an incorrect draft from becoming an external communication.

The workflow compares freight invoices with agreed rates, shipment details and warehouse records. It prepares a discrepancy workbook and carries open queries forward, giving the reviewer the information needed to investigate mismatches.

A discrepancy is presented for review, not treated as a confirmed error. The invoice is checked against supporting operational records, and the team decides what to challenge or approve instead of relying on an AI conclusion alone.

A scheduled workflow collects promotion information from emails and attachments, identifies the relevant year, month and week, and updates the shared workbook. References are retained so the sales team can trace the promotion back to its source.

The workflow targets the agreed workbook and includes duplicate checks. Keeping the original references alongside the extracted information makes an incorrect date or promotion easier to identify and correct.

The cockpit gathers inventory and demand information, combines it with ingredient and production constraints, and includes packaging calculations. The team starts its scheduling work with the relevant inputs together rather than assembling separate exports.

The cockpit supports the planning decision; the operations team makes it. Spreadsheet cross-checks are retained so calculations can be compared before the team commits to a production plan.

The assistant collects purchase orders from incoming communications and brings them together with the related work information. It helps identify incomplete, outdated or unresolved orders and presents the information needed for reconciliation and follow-up.

The workflow is semi-automated: the person responsible reviews the prepared information and makes the decision. Collecting and checking an order is kept separate from approving a payment or committing the business.

The workflow starts from supplied business bank exports, merges transactions and removes duplicates using bank identifiers. It matches supporting invoices to real transactions and prepares the reconciliation and close report. Missing proof is surfaced in the working file so the reviewer can resolve it before closing the period.

Amounts come from the bank records, with personal banking kept separate. Supporting files must pass the invoice verification check before the close can proceed. The accountant can review the transactions and evidence instead of relying on an unsupported match.

The workflow reads product catalogues, groups product families and variants and organises the scope for a quote. A pricing snapshot supplies the agreed rates to the calculator, so the arithmetic follows a defined basis rather than being improvised in a conversation. The person preparing the quote reviews the resulting scope and price.

Uncertain product information is flagged for review. The calculator requires the pricing inputs instead of silently substituting zero for a missing rate. Product selection and the final commercial commitment remain with the person approving the quote.

The assistant retrieves purchase information, amounts, locations and receipt photos, then enters the information into the team’s shared spreadsheet. The finance team has the transaction details together instead of collecting them separately.

Receipt information remains part of the record available to the team. Capturing an expense in a spreadsheet is distinct from authorising spending or approving reimbursement, keeping those decisions with the responsible people.

The workflow reviews overdue accounts alongside payments, credits and previous correspondence. It prepares the customer information and, where appropriate, a draft demand letter for the finance team to review.

Letters remain drafts for human review and approval. The workflow does not send them automatically or change the accounting system, helping prevent an unchecked balance from triggering an incorrect customer demand.

The workflow brings budget holders, accounts and leadership into a structured request and approval process. Requests, decisions and follow-up are organised together so people can see where a spending decision sits.

Approval remains with the designated people in the process. A submitted request is kept distinct from an approved spend, reducing the risk of someone treating a request or an AI-prepared entry as permission to proceed.

The dashboard brings leadership information into a shared view of priorities, performance and follow-up. It gives decision-makers a consistent place to review the business information that would otherwise be spread across separate updates.

The dashboard provides visibility for leadership decisions. It is separate from the budget approval process: seeing a figure or a priority in the dashboard does not authorise a payment or replace the responsible person’s decision.

The workflow applies the target customer criteria, removes unsuitable businesses and duplicates, and organises prospects for qualification. The user can keep, reject or choose a manual contact route for each prospect. Email verification then helps separate usable addresses from contacts that need a different approach.

Research does not trigger outreach. Existing pipeline exclusions and deduplication reduce repeat contact, while uncertain email results are kept out of the verified email list and routed for manual handling where appropriate. The user reviews who should be contacted.

The workflow prepares initial emails and follow-ups around the target customer profile, supports A/B campaigns and tracks campaign performance. Prospect lists are replenished at defined thresholds, while campaign rules manage when activity should stop.

Campaign activity follows configured rules, including stop thresholds and email warm-up. Those limits give the campaign a defined operating boundary instead of allowing it to run indefinitely without reference to its results.

The assistant combines trend research, supplied screenshots and available content to prepare posts and visual assets. Draft content is organised in a review folder; once approved, it can be scheduled to the business’s social accounts.

Human validation sits between content preparation and scheduling. The person reviews the material before it goes into the publishing queue, helping prevent an unsuitable draft or generated asset from being published unchecked.

Marketing assets are organised into a searchable collection, making the files needed for recurring work easier to locate. The team can retrieve existing material rather than searching through disconnected folders or rebuilding assets.

The library helps people find existing source material. Selecting and checking the right asset remains part of the team’s work; retrieving a file does not publish it or establish that it is appropriate for every campaign.

The setup captures transcripts, important communications and working knowledge in one place. That material supplies context to the AI used for later tasks, reducing the need to explain the same decisions and background repeatedly.

The captured material provides working context, not automatic authority to act. Sensitive actions still follow the agreed human approval boundaries, and session recording requires the participant’s agreement under our data-handling approach.

Projects, context and memory files are organised around the person’s role and recurring tasks. The assistant can draw on relevant working information across sessions, helping the person continue useful work instead of restarting the explanation each time.

Context and access are scoped to the agreed work. The assistant’s knowledge of a project does not itself grant permission to send messages, change systems or make commitments; those actions follow the access and approval boundaries set for the work.

The assistant helps organise design references and translate visual ideas into draft assets. It supports the preparation work around marketing and client communication, making it easier to move from an idea to material that can be reviewed.

Draft assets are working material for human selection and review. The person using them checks their suitability for the intended purpose before they become client-facing content or part of a publication.

The workflow sets up cameras, lighting and a sequence of angles in Blender, then renders the product views in a batch. Reusable scene settings keep the framing consistent across a range, while the operator can adjust materials and lighting before generating the final images.

The source model is kept intact. Test renders let the operator check framing, materials and colour before committing to the full batch. Rendered outputs are reviewed before delivery so an incorrect scene setting does not silently carry through to the client.

Scripts handle repeatable image preparation around a product shoot, including shadow cleanup, colour checks against different backgrounds and web optimisation. The resulting files can be checked together, making it easier to spot inconsistent colour or unwanted edges across a set of product views.

Original source files are preserved. Colour and edge checks provide concrete points for review, and the operator inspects the output before it enters the delivery set. Changes can be rerun from the source if an adjustment damages the product’s appearance.

The workflow reviews the model and its materials, prepares compression candidates and compares the resulting files. Material-specific settings allow different parts of a product to be treated appropriately, rather than applying one compression setting to every texture and accepting the result unchecked.

The original asset remains unchanged. Candidate files are inspected before selection, with appearance and model integrity checked alongside size. A smaller file is only useful if it still represents the product correctly.

The setup workflow reads the model’s parts and dimensions and guides the preparation of its viewer. The operator places hotspots, supplies labels and selects the relevant viewing or AR options. Calculated dimensions provide a starting point for confirmation instead of requiring every value to be entered again.

Dimensions and labels are reviewed before use. The model is checked in staging before the live viewer is updated, helping catch missing parts, incorrect measurements or a configuration that behaves differently from the intended product experience.

The workflow checks incoming files, identifies missing product information and connects the technical setup with the business delivery record. It brings model preparation, configuration and delivery checks into one repeatable sequence so each product has a traceable path from intake to release.

Source files are preserved and missing information is surfaced for resolution. Staging and validation separate preparation from release, so a newly processed asset does not become a live client deliverable simply because a script finished.

At the end of a session, the workflow captures the relevant decisions, corrections and next steps in the project’s retained context. Later sessions can load those records, giving the assistant a working starting point without asking the person to explain the same history again.

Retained context is stored in controlled project files with version history and backups. Current source records take precedence over older notes, and important facts are checked before action so an outdated summary does not become an instruction to repeat a mistake.

This internal dashboard was built to connect business records for tasks, goals, clients, procedures, quotes and deliverables. It brought the same operating information into connected views, with supported assistant updates. Day-to-day task management has since moved to ClickUp; this remains an example of a tool we built.

Access is authenticated and the assistant’s supported actions are limited. Keeping changes tied to the underlying business records makes them reviewable, while restricting the available operations reduces the scope of an incorrect request.

The workflow records a guided page scroll, assembles the personalised video and pairs the uploaded result with the intended prospect. A tracking record brings together the video link and available viewing information, making it easier to prepare a batch without manually rebuilding every demonstration.

The prospect and video pairing is checked before outreach. Preparing and uploading the video is separate from sending it, leaving a human review step to catch the wrong name, page or demonstration before a prospect receives the message.

The assistant reviews campaign exports and incoming responses alongside the outreach context. It helps distinguish interest, objections and messages that need a different follow-up, then prepares the next step for review. The owner can decide how to respond without rereading the entire campaign history.

Suggested next steps remain drafts for the owner’s decision. A reply does not automatically trigger another message or change a commercial commitment, keeping interpretation errors from turning into an unwanted follow-up.

The assistant works from the owner’s examples, writing preferences and previous corrections to prepare new drafts. Content can be organised by angle and audience, helping the owner develop ideas consistently while retaining the specific language and points they want to make.

The owner reviews the content before publication. Retained corrections guide the writing, while claims and examples still need support from the source material. A stylistically convincing draft is not treated as evidence that its statements are true.