Claims that artificial intelligence will “save thousands” are easy to make and often impossible to audit. A useful business case starts somewhere less exciting: current transaction volumes, minutes per task, employment cost, technology cost and the percentage of work that can realistically be assisted.

This article provides an illustrative UK model for September 2026. It is not a guaranteed quotation or a recommendation to remove roles. It shows how to translate saved time into gross capacity, then subtract licensing, implementation and support to estimate a more credible annual benefit.

Begin with an employment-cost baseline

From 1 April 2026, the UK National Living Wage for workers aged 21 and over is £12.71 per hour. At 37.5 hours per week for 52 weeks, that is £24,784.50 in gross annual wages. The official rate is published by the UK Government.

Wages are not the entire employer cost. For the 2026–27 tax year, the standard employer National Insurance rate is 15% above the £5,000 annual secondary threshold. On £24,784.50, that produces approximately £2,967.68 of employer National Insurance before considering reliefs such as Employment Allowance. Employers must also contribute at least 3% of qualifying earnings to an eligible workplace pension. The relevant current figures are in HMRC’s 2026–27 employer rates and the government’s pension guidance.

Using a simplified pension calculation of 3% on earnings between £6,240 and £24,784.50 adds about £556.34. The direct annual employment cost in this example is therefore approximately £28,308.52, or £14.52 per paid hour. Holiday cover, equipment, office space, recruitment, training and management time would increase the fully loaded cost, but they are deliberately excluded from this conservative baseline.

Scenario one: an AI receptionist

An AI receptionist can answer routine calls, capture details, provide approved information, qualify enquiries, book appointments and transfer urgent or complex conversations to a person. The business case should not assume that every call is automated. Customers may have accessibility needs, emotional situations, unusual questions or a clear preference for a person.

Assume a business receives 50 routine calls per working day. The average call, note-taking and follow-up take four minutes, creating roughly 1,000 minutes or 16.7 hours of weekly handling. If the AI receptionist safely completes 60% of those interactions, it releases about 10 hours per week.

At the conservative £14.52 direct hourly employment cost calculated above, 520 hours represent £7,550 of gross annual capacity. If the receptionist platform, telephony, setup and support cost £4,200 in year one, the illustrative first-year benefit is £3,350. If the volume is higher, the handling time is longer or the released capacity avoids overtime or an additional hire, the value can be substantially greater. If call volume is low, the business case may not work at all.

Scenario two: an AI chatbot and enquiry triage

A website or internal chatbot can answer questions grounded in approved content, collect structured enquiry details and route a conversation to the correct team. The useful measure is not “messages sent”. It is the reduction in repeat handling while maintaining customer satisfaction and escalation quality.

Suppose two team members collectively spend eight hours per week answering repeated questions, requesting missing details and forwarding enquiries. At an illustrative loaded cost of £20 per hour, that work represents £8,320 annually. If a well-designed chatbot and routing process reduce the effort by half, the gross capacity released is £4,160. After £1,800 in annual platform and support costs, the illustrative net benefit is £2,360.

The chatbot should make its boundaries clear, use controlled source material, avoid inventing answers and provide a visible route to a person. A weak chatbot can increase costs by creating confused customers and additional complaint handling.

Scenario three: AI-assisted design and content production

AI design tools can create first drafts, resize approved assets, remove backgrounds, prepare variations and support repetitive production. Adobe’s Photoshop API documentation describes automated background removal, Smart Object replacement, text editing, presets and Photoshop Actions at scale. That makes the savings opportunity more concrete than a vague promise that AI is “creative”. See Adobe’s Photoshop automation examples.

Assume a marketing team spends six hours each week adapting approved creative into different formats and preparing routine variants. At £30 per hour, the annual labour value is £9,360. If automation removes 50% of the production handling while designers retain concept, brand and final approval, the gross capacity released is £4,680. After £1,500 of incremental tools, templates and support, the illustrative annual benefit is £3,180.

Scenario four: AI-assisted coding

Coding assistance can generate boilerplate, explain unfamiliar code, draft tests and suggest changes. The saving depends heavily on task type, code quality, review discipline and developer experience. Generated code still requires security review, testing and accountable ownership.

Microsoft’s AI and Productivity Report describes a controlled GitHub Copilot study involving 95 developers completing an HTTP server task. The report discusses speed as output per unit of time and forms part of a wider evidence base on generative AI productivity. Microsoft’s separate workplace experiment found people using Copilot completed searching, writing and summarising tasks 29% faster. These results are useful evidence, but neither percentage should be applied mechanically to every development team. See the Microsoft AI and Productivity Report.

For a cautious example, assume a developer spends five hours per week on suitable repetitive work at £45 per hour. That is £11,700 annually. If assistance releases 25% of that time, the gross capacity is £2,925. After £600 in tools and governance, the illustrative annual benefit is £2,325. The larger benefit may be faster delivery and fewer queues rather than a reduction in headcount.

Scenario five: Microsoft 365 administration automation

This is often the largest opportunity because administration is distributed across many employees. Consider 15 people who each spend two hours per week copying information, chasing approvals, renaming files, preparing routine documents and updating colleagues. At a loaded cost of £25 per hour, that work represents £39,000 annually.

If SharePoint, Power Apps, Power Automate and Copilot remove 50% of the effort, the gross capacity released is £19,500. Allow £9,000 in the first year for discovery, configuration, testing, training, licensing and support. The illustrative first-year benefit is £10,500, rising to £15,500 in a later year if ongoing licensing and support are £4,000.

Microsoft cites a Forrester Power Apps study in which line-of-business employees improved productivity by 3.2 hours per week, application-development costs fell by 74% and three-year ROI reached 188% in the composite organisation. A later Microsoft summary cites up to 250 hours saved per user annually for high-impact use cases. These are vendor-sponsored composite findings, not automatic outcomes. They should inform a range, not replace a business-specific baseline. Review Microsoft’s Power Apps economic-impact summary.

A combined illustrative annual model

In this model, a business releases £38,815 of gross annual capacity and spends £17,100 on technology, implementation and support, producing an illustrative first-year benefit of £21,715. The gross figure is not the same as cash in the bank. Real financial value appears when the organisation converts capacity into measurable outcomes: additional work completed, overtime avoided, faster invoicing, lower outsourcing spend, fewer errors or a hire deferred.

How to calculate your own number

For each process, record annual volume, average handling time and realistic loaded hourly cost. Multiply them to find the current annual cost. Estimate the percentage that automation can safely remove, using a pilot rather than optimism. Then subtract annual licences, telephony or usage charges, implementation, maintenance, monitoring and staff training.

Run at least three cases: conservative, expected and high-volume. Include exception handling and human review. If the expected case depends on 100% automation or zero ongoing support, it is not an expected case.

Finally, measure after launch. A credible automation programme keeps a short benefits register showing transaction volume, hours avoided, failure rate, escalation rate, customer satisfaction and the business use made of released capacity.

The purpose is better allocation, not simply fewer people

Receptionists understand context, reassure customers and handle exceptions. Designers make judgements about meaning and brand. Developers own architecture, quality and security. Administrators notice the unusual case that a workflow designer did not anticipate. Automation is most valuable when it removes avoidable repetition and gives those people more time for the work that requires judgement.

The honest answer to “how much can AI save?” is therefore a range produced from your own operational evidence. For a small organisation, a controlled combination of reception automation, enquiry triage, creative assistance and Microsoft 365 workflows could plausibly release tens of thousands of pounds in annual capacity. Whether that becomes a true saving depends on adoption, process quality and what the organisation does with the time.

Technology, made practical

Designed around real people and real work.

Agentic AI changes the role of artificial intelligence from answering a question to completing a sequence of actions. Given a defined objective, an agent can observe a computer, decide the next step, use a mouse and keyboard, check the result and continue. That makes it possible to work with software designed for people, including specialist applications that do not offer a complete modern API.

This does not mean handing an unsupervised robot unrestricted control of a production computer. Useful computer automation combines AI reasoning with boundaries: approved applications, limited files, staged outputs, audit logs and human confirmation before consequential actions.

What computer-using AI actually does

A computer-using agent normally operates as a loop. It receives a goal and contextual rules. It observes a screenshot or structured representation of the interface. It reasons about the next action, then clicks, types, scrolls or invokes a tool. After the interface changes, it observes again and checks whether the action had the intended effect.

OpenAI describes its Computer-Using Agent as combining visual understanding with reasoning so it can interact with graphical interfaces through mouse and keyboard actions. The published research also emphasises limitations, confirmation before external side effects and defences against malicious instructions embedded in webpages. Those safeguards are not obstacles to deployment; they are part of a responsible design. See OpenAI’s explanation of the computer-using agent.

There are three practical ways for an agent to control software:

• Direct interface control: the agent uses the visible application like a person. This is flexible but can be affected by layout changes, pop-ups and ambiguous screen states.
• Native scripting or APIs: the agent generates and runs structured commands through the application’s supported automation layer. This is usually faster, repeatable and easier to test.
• Hybrid control: the agent uses an API or script for precise production work and the graphical interface for setup, review and functions that are not exposed programmatically.

The hybrid model is often the strongest. It allows AI to understand an outcome in human language while using deterministic tools for the parts that must be exact.

Blender: from a brief to a repeatable 3D workflow

Blender is particularly suitable because its embedded Python environment exposes scene data and operations. Its official documentation lists automation, scene manipulation, import, export, object creation and rendering among typical scripting uses. An agent can therefore translate a structured brief into a Python script that creates objects, applies materials, positions cameras, sets render options and exports approved formats.

For a product-visualisation workflow, the agent might read a product specification, create a base scene from a controlled template, import the correct model, apply named materials, generate three camera views and render preview images. A human designer reviews the previews before high-resolution rendering. This is safer than asking an agent to improvise every click and gives the studio a script it can rerun. Blender’s current documentation explains the embedded Python API and automation model.

Canva and browser-based design tools

Browser-based design software such as Canva can be controlled through a combination of templates, data-driven creation features, platform integrations and supervised interface actions. The most dependable workflow starts from an approved brand template. The agent selects the correct format, inserts supplied copy and imagery, checks required fields and creates a draft for review.

It should not independently invent legal claims, prices or brand assets. It should not publish directly to social channels without confirmation. The valuable automation is the repetitive production work: creating correctly sized variants, replacing campaign details, assembling a first draft and naming files consistently.

Adobe Photoshop: actions, scripts and cloud APIs

Photoshop offers stronger automation routes than many people realise. Adobe documents UXP JavaScript scripting, plug-ins and cloud APIs, including Photoshop Actions. Its APIs can automate image processing, background removal, lighting adjustments, Smart Object replacement, text-layer edits and large batches of renditions.

An agent can decide which approved workflow applies, validate the input and call the appropriate action or API. For an e-commerce catalogue, that might mean isolating a product, applying a consistent crop, placing it into a Smart Object template and exporting web, marketplace and social sizes. The agent then checks dimensions and file names before handing the results to a designer. Adobe’s developer documentation describes both Photoshop scripting and Actions and the current Photoshop API.

Music production and video editing

Digital audio workstations and video editors contain large amounts of repeatable operational work. An agent can create project folders, import named assets, align them to a template, label tracks, place markers, organise takes, prepare proxy media, generate captions, assemble rough selects and export review versions.

Creative judgement still belongs to the producer or editor. An AI-generated rough cut can reduce the time spent locating material, but pacing, emotional intent, music rights and final quality need accountable human review. In music production, the agent can prepare routing, file organisation and repetitive edits, while the producer makes decisions about performance, arrangement, sound and artistic direction.

The best design separates reversible preparation from consequential output. Renaming copied files is low risk. Overwriting a master recording is not. Creating a preview render is low risk. Publishing the final video or sending stems to a client requires confirmation.

Our Asta Powerproject documentation workflow

We applied the same principles for a customer that needed documentation around Asta Powerproject. The challenge was not simply writing text. The process required moving between specialist scheduling software, source project information, screenshots, structured instructions and consistent document formatting.

The agent-supported workflow began with an agreed document outline and a controlled set of example project files. The computer-using layer navigated the software to the relevant views and commands, while the documentation layer recorded the task, the expected result and any prerequisite. Screenshots were captured at defined checkpoints rather than whenever the screen happened to look useful. Draft instructions were then assembled into a consistent structure for human review.

Asta Powerproject supports task, resource and cost reporting, including tabular reports and Business Intelligence export to Excel. Its documentation describes reports for upcoming tasks, critical work, variance, resources and costs. That made it possible to organise guidance around real outputs rather than a tour of every menu. See Elecosoft’s documentation on producing reports and report and Business Intelligence options.

Human checking remained essential. A knowledgeable reviewer confirmed that each instruction matched the customer’s version and working practice, that screenshots did not expose confidential project data and that terminology was consistent. The agent reduced repetitive capture, drafting and formatting work; it did not become the technical authority.

Guardrails for computer control

A responsible implementation should use a dedicated environment with the minimum permissions required. Source files should be read-only where possible, and outputs should go to a separate review folder. Approved scripts and templates should be version-controlled. Actions that publish, overwrite, purchase, message or change access should require a person to confirm.

Every workflow also needs a stopping rule. If the screen is different from the expected state, a file is missing or the output fails validation, the agent should pause and report the problem. Quietly improvising around uncertainty is precisely what a production automation should avoid.

Where agentic AI is most valuable

The strongest candidates combine clear inputs, repeated steps and reviewable outputs. Examples include producing consistent software documentation, creating batches of branded assets, setting up 3D scenes, preparing media projects and converting structured data into reports. The weakest candidates are ambiguous one-off tasks where success is subjective and errors are hard to detect.

Computer-using AI widens the range of software that can participate in automation. Native APIs and scripts should still be preferred for precision. The graphical interface becomes a bridge for long-tail tasks and legacy systems, with the agent acting as an adaptable operator under defined supervision.

Technology, made practical

Designed around real people and real work.

Most organisations do not suffer from a shortage of software. They suffer from disconnected information, repeated data entry and processes that depend on somebody remembering the next step. Microsoft 365 automation can change that by connecting tools many businesses already license: SharePoint, Microsoft Lists, Power Apps, Power Automate, Teams, Outlook and Copilot.

The important word is connecting. A form on its own does not transform a process. A flow that sends an email is not automatically a dependable system. The value appears when information has one controlled home, staff receive a simple interface, routine decisions move automatically and exceptions reach the right person with enough context to act.

A realistic example: automating a service request

Imagine a property, construction or professional-services business that receives internal requests for purchases, site visits, subcontractor documents and customer work. The existing process begins with an email. Details are copied into a spreadsheet. Someone creates a folder, asks for approval, chases missing information, prepares a confirmation document and sends progress updates. At month end, a manager tries to reconcile the spreadsheet with messages and invoices.

None of those steps is individually difficult. Together they create hours of invisible administration. They also create risk: duplicate records, inconsistent file names, unclear approval status and decisions buried in inboxes.

A connected Microsoft 365 solution can replace the loose chain with one understandable process:

1. A Power Apps form collects the request using conditional questions and validation.
2. The record is written to a controlled SharePoint or Microsoft List.
3. Power Automate creates the correct folder structure, applies metadata and starts the appropriate approval route.
4. Approvers receive a structured request in Teams or Outlook and can approve, reject or request clarification.
5. Copilot helps summarise supporting material or draft a consistent customer-facing response.
6. The decision, comments and dates are written back to the record, creating an audit trail.
7. Dashboards show volume, status, bottlenecks and approaching deadlines without another manually maintained spreadsheet.

SharePoint provides the controlled information foundation

SharePoint is more than a place to store Word documents. Lists can hold structured records, while document libraries manage files, versions, permissions and metadata. A request can therefore have an owner, status, department, value, due date and approval history alongside its supporting documents.

This matters because automation needs reliable data. If every team uses a different spreadsheet or naming convention, a flow has to guess. If the organisation agrees a small number of fields and ownership rules, the process becomes measurable. Permissions can reflect real responsibilities, and retention or sharing controls can be applied consistently.

Microsoft documents more than 100 SharePoint templates for Power Automate and identifies approvals, file movement, permissions and reminders as common scenarios. Its guidance also shows how document approval can trigger when a file is added, send the request to an approver, update the document’s approval status and notify the author. See Microsoft’s guidance on SharePoint workflows and automated document approval.

Power Apps gives staff a cleaner way to work

People should not need to understand the database behind a process. Power Apps can provide a focused phone, tablet or desktop interface that shows only the questions and actions relevant to the user. A site manager might create a request and photograph supporting evidence. Finance might see cost codes and approval status. A director might see items requiring a decision.

Microsoft supports creating a responsive canvas app directly from a SharePoint or Microsoft List. The resulting app can browse, add, edit and manage list records, while changes made through the app update the underlying SharePoint data. That gives a small business a practical route from spreadsheet administration to a governed application without commissioning a large bespoke system. Microsoft’s current walkthrough is available in its Power Apps and SharePoint guide.

Power Automate becomes the process engine

Power Automate watches for events and performs the repeatable work. It can react when a list item is created, an email arrives, a form is submitted, a date approaches or a document changes. It can look up information, apply conditions, request approvals, create files, post to Teams, update records and call other systems through connectors or APIs.

In the service-request example, the flow should do more than follow the happy path. It needs to handle missing approvers, rejected requests, resubmission, duplicate references and failures when an external system is unavailable. Each automated step should have an owner, a meaningful error message and a route for human intervention.

Microsoft’s approval framework can manage documents and processes across SharePoint, OneDrive, Dynamics 365 and other connected services. An approval can wait for a response, record comments and update the originating record. That is far more dependable than forwarding an email and hoping the decision is captured later. See Microsoft’s approval workflow documentation.

Copilot handles the information-heavy parts

Traditional automation is strongest when the rules are explicit. Generative AI is useful where the input is less structured: a long email, meeting transcript, survey response, report or set of project notes. Copilot can help summarise, classify and draft, while Power Automate controls when the task happens and where the result goes.

For example, Copilot could produce a concise briefing for an approver from a request and its attachments. It could draft a customer update using approved facts from the SharePoint record. It could turn meeting notes into proposed actions. The output should still be reviewed where accuracy, commitments or sensitive information matter.

Microsoft reported that participants using Copilot were 29% faster across a set of searching, writing and summarising tasks: 29 minutes 42 seconds with Copilot compared with 42 minutes 6 seconds without it. The experiment does not guarantee a 29% saving in every workplace, but it demonstrates why information-heavy administration is a sensible area to test. Read the Microsoft Work Trend Index research.

What could the saving look like?

Consider a team of 12 people who each spend two hours per week entering information, chasing approvals, moving documents and producing routine updates. That is 1,248 hours per year. At an illustrative loaded employment cost of £28 per hour, the activity represents £34,944 of annual capacity.

If a well-designed solution removes 60% of that effort, the gross capacity released is about 749 hours, worth approximately £20,966 per year. This is not automatically a cash saving and should not be presented as one. The organisation may use the time to serve more customers, shorten response times, reduce overtime or avoid adding another administrative role as volume grows. Implementation, licensing, support and training costs must be deducted before calculating return on investment.

Microsoft cites a Forrester study in which high-impact Power Apps use cases saved up to 250 hours per user annually, while another Microsoft summary reports an average 3.2 hours of line-of-business productivity improvement per user per week. Those are composite-study findings rather than a promise for a specific company, but they provide useful benchmarks when building a cautious business case. See Microsoft’s summaries of the Power Apps economic-impact study and reported productivity gains.

A sensible delivery sequence

Start with one process that is frequent, visible and irritating. Measure the current number of transactions, handling time, error rate, waiting time and rework. Map the real exceptions before choosing technology. Then build the smallest end-to-end version, test it with real users and compare the result with the baseline.

A useful first phase normally includes a SharePoint information structure, a focused Power Apps interface, one or two Power Automate flows, clear ownership and a simple reporting view. Copilot should be introduced where it adds measurable value, not attached to every step for novelty.

The final result is not “an app” or “a flow”. It is a managed way of working: people know where to start, the next action reaches the right person, decisions are recorded and managers can see the process without asking someone to rebuild the story in Excel.

Technology, made practical

Designed around real people and real work.

Welcome to WordPress. This is your first post. Edit or delete it, then start writing!

Technology, made practical

Designed around real people and real work.