{"id":4448,"date":"2026-11-02T17:01:55","date_gmt":"2026-11-02T13:01:55","guid":{"rendered":"https:\/\/artinwebs.com\/blog\/which-process-to-automate-first\/"},"modified":"2026-11-02T17:02:46","modified_gmt":"2026-11-02T13:02:46","slug":"which-process-to-automate-first","status":"publish","type":"post","link":"https:\/\/artinwebs.com\/blog\/which-process-to-automate-first\/","title":{"rendered":"Which Process to Automate First: How to Compare and Rank Your Candidates"},"content":{"rendered":"<p><span>To decide which process to automate first, list your candidates, rule out the ones that fail a basic suitability check, then score the rest on two things kept separate, how much time and trouble automation would save and how hard it would be to build. The first project should do well on both and be cheap to get wrong.<\/span><\/p>\n<p><span>This guide picks up where <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/artinwebs.com\/blog\/business-automation-where-to-start\/&amp;sa=D&amp;source=editors&amp;ust=1793628034790888&amp;usg=AOvVaw0kGZWCC-Lu_G2eFysILtDd\">the plan for starting business automation<\/a><\/span><span>&nbsp;leaves off. That guide explains why the first project should be a single process. This one shows how to compare several candidates and choose between them, with a scoring sheet you can copy. The criteria come from two published methods, the discover stage of the NSW Government&#8217;s Automation Guide and the assessment model in UiPath&#8217;s Automation Hub documentation. Both were read on 2 November 2026.<\/span><\/p>\n<h2 id=\"h.kaj2shdadn6h\"><span>Collect the candidates from the people who do the work<\/span><\/h2>\n<p><span>Start with a list, not a favourite. The people who do a task every day know how often it really runs, how long it takes and where it goes wrong, and those three facts decide most of the ranking. The <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/www.digital.nsw.gov.au\/delivery\/nsw-automation-guide\/automation-lifecycle\/stage-1-discover&amp;sa=D&amp;source=editors&amp;ust=1793628034791990&amp;usg=AOvVaw09Oae3kUUt3OaPd97DFmHy\" target=\"_blank\" rel=\"noopener\">discover stage of the NSW Automation Guide<\/a><\/span><span>&nbsp;gives the same advice to larger organisations, telling them to &#8220;involve relevant multidisciplinary stakeholders when identifying and prioritising automation tasks.&#8221;<\/span><\/p>\n<p><span>For each candidate, write one line that says what starts it, how often it runs, who does it and which systems or files it touches. You will need those details for every step that follows, so it is worth getting them right now rather than guessing later.<\/span><\/p>\n<h2 id=\"h.1opir7i8ecuo\"><span>Which processes should you rule out first?<\/span><\/h2>\n<p><span>Before scoring anything, run each candidate through a quick yes or no check. The NSW guide publishes a task suitability checklist, and a shortened version of it works for a small business.<\/span><\/p>\n<ul>\n<li><span>Is the work mostly manual and repetitive?<\/span><\/li>\n<li><span>Does it follow rules that are clear and stay the same from week to week?<\/span><\/li>\n<li><span>Are exceptions rare?<\/span><\/li>\n<li><span>Does the input arrive in a standard digital form, such as a web form, a spreadsheet row or an email with a fixed layout?<\/span><\/li>\n<li><span>Will the systems it touches stay in place for the foreseeable future?<\/span><\/li>\n<li><span>Can you measure it today, in runs per week and minutes per run?<\/span><\/li>\n<\/ul>\n<p><span>A candidate with two or more &#8220;no&#8221; answers belongs on a later list, not in the bin. Often the process can be changed so that it passes, for example by replacing free-text emails with a short form, and it becomes a good candidate once that change has settled.<\/span><\/p>\n<p><span>The NSW checklist also sets a size threshold, asking whether &#8220;the task when automated saves the time of at least 2 to 3 full-time employees (FTE).&#8221; That bar suits a large organisation. A small company can set a lower one, but it should decide the number before scoring starts, so that nobody&#8217;s favourite project gets through on enthusiasm alone.<\/span><\/p>\n<p><span>Two more questions from the same guide do not rule a process out, but they make it a harder first project. One asks &#8220;does the task or process involve handling sensitive or confidential data?&#8221; and the other asks you to &#8220;decide if the task requires human intervention or decision-making.&#8221; If the answer to either is yes, the process can still be automated later, with more review and testing than a first project usually gets.<\/span><\/p>\n<h2 id=\"h.u8zquomeyctz\"><span>How to estimate what each candidate is worth<\/span><\/h2>\n<p><span>Value comes from three places. Staff time saved, errors avoided and delay removed for whoever waits on the result, whether that is a customer or the next team. Time is the easiest to count. Multiply the runs per year by the minutes per run and you have the hours the process takes today.<\/span><\/p>\n<p><span>Not all of those hours will be saved, because some runs are exceptions that stay with a person. UiPath&#8217;s <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/docs.uipath.com\/automation-hub\/automation-cloud\/latest\/user-guide\/information-about-the-detailed-assessment-algorithm&amp;sa=D&amp;source=editors&amp;ust=1793628034795362&amp;usg=AOvVaw22LLal5GICMCpSWGoxE9S6\" target=\"_blank\" rel=\"noopener\">detailed assessment algorithm<\/a><\/span><span>&nbsp;handles this with a score it calls Automation Potential, described as &#8220;a measure of the process\/task automation suitability,&#8221; running from 0 to 100 percent. Its hours benefit is that percentage multiplied by the total time the process takes today. You can use the same idea without the software by estimating the share of runs that follow the common path and counting only those.<\/span><\/p>\n<p><span>Here is an invented example to show the arithmetic. A task that runs 400 times a month at 6 minutes a run takes 40 hours a month. If one run in four is an exception that a person still handles, the realistic saving is about 30 hours a month, before you subtract the time someone spends checking the automation.<\/span><\/p>\n<p><span>For errors, note how often a run has to be corrected and what the correction costs, including any complaint or credit note that follows. For delay, note who is waiting and for how long. A process that saves few hours can still rank high if its mistakes reach customers or hold up payment.<\/span><\/p>\n<p><span>For a quick sense check on frequency, the xkcd comic <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/xkcd.com\/1205&amp;sa=D&amp;source=editors&amp;ust=1793628034797141&amp;usg=AOvVaw0uRPD688gzaxL0huyXuqSk\" target=\"_blank\" rel=\"noopener\">Is It Worth the Time?<\/a><\/span><span>&nbsp;is a grid of how long you can spend making a routine task more efficient before you spend more time than you save across five years. As the <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/www.explainxkcd.com\/wiki\/index.php\/1205&amp;sa=D&amp;source=editors&amp;ust=1793628034797440&amp;usg=AOvVaw06ALRkrD3eUr4ZfaE3Khsh\" target=\"_blank\" rel=\"noopener\">explain xkcd page<\/a><\/span><span>&nbsp;points out, the chart treats all time as equal and assumes only one person benefits, so use it to spot obvious mismatches, not as a business case.<\/span><\/p>\n<h2 id=\"h.1cck0vo5vbhp\"><span>How to judge how hard each one will be to build<\/span><\/h2>\n<p><span>The second score is effort. UiPath calls it Ease of Implementation and lists six factors behind it, namely process stability, application stability, how structured the input data is, process variability, process length and the number of applications used. It also flags scanned input that needs character recognition. The result is banded as easy from 65 to 100 percent, medium from 35 to 65 percent and difficult below 35 percent.<\/span><\/p>\n<p><span>You do not need the tool to use the factors. For each candidate, ask these questions.<\/span><\/p>\n<ul>\n<li><span>How often has the process itself changed in the past year?<\/span><\/li>\n<li><span>Do the applications it uses change their screens or data formats often, and do they offer a built-in integration or an API?<\/span><\/li>\n<li><span>Is the input structured, as form fields or spreadsheet columns, or is it free text, PDF files or scans?<\/span><\/li>\n<li><span>How many versions of the process exist, for example different rules for different customer types?<\/span><\/li>\n<li><span>How many steps does it have and how many separate applications does it touch?<\/span><\/li>\n<\/ul>\n<p><span>The NSW guide adds data preparation and integration with existing systems to its list of context factors. Both tend to show up as effort. If customer names are spelled three ways across two systems, someone has to decide which one is right before any automation can match them.<\/span><\/p>\n<figure><img decoding=\"async\" src=\"https:\/\/artinwebs.com\/blog\/wp-content\/uploads\/2026\/11\/which-process-to-automate-first-value-ease-awg-mvz9px45.jpg\" alt=\"Chart with process cards placed by value and ease, the high-value, easy card highlighted green\" title=\"Value and ease side by side\"><figcaption>Candidates that are high on value and easy to build go to the short list first.<\/figcaption><\/figure>\n<h2 id=\"h.8u7og8o6rsm\"><span>A scoring sheet you can copy<\/span><\/h2>\n<p><span>Once value and effort are estimated, put the candidates side by side. A simple sheet is enough.<\/span><\/p>\n<ol start=\"1\">\n<li><span>Put each candidate that passed the suitability check on its own row.<\/span><\/li>\n<li><span>Score value from 1 to 3 on hours saved, errors avoided and delay removed, and add the three together.<\/span><\/li>\n<li><span>Score ease from 1 to 3 on stability, input, number of variations and number of systems, where 3 is the easiest, and add those together.<\/span><\/li>\n<li><span>Score the cost of a mistake from 1 to 3, where 3 means a wrong result stays inside the company and is easy to spot and undo.<\/span><\/li>\n<li><span>Keep the three totals in separate columns rather than multiplying them into one number, so you can see why each candidate ranks where it does.<\/span><\/li>\n<li><span>Short-list the candidates that are high on both value and ease, then pick the one whose mistakes are cheapest.<\/span><\/li>\n<\/ol>\n<p><span>Useful columns are the candidate&#8217;s name, runs per month, minutes per run, the share of runs on the common path, the three scores, the person who owns the process and a note on its exceptions. The owner column matters more than it looks. A candidate nobody can sign off on will stall at the testing stage, whatever its scores.<\/span><\/p>\n<p><span>Here is a second invented example, not taken from any client, to show how the sheet reads. A company has three candidates. Copying web-form enquiries into the CRM has steady volume, structured input and two systems, and a wrong entry is easy to fix, so it scores medium on value, high on ease and high on safety. Matching supplier invoices to purchase orders saves the most time, but the invoices arrive as PDF files in many layouts, so it scores high on value and low on ease. Replying to customers about order status is valuable, but a wrong answer goes straight to a customer, so it scores low on safety. The form-to-CRM job is the better first project, and invoice matching becomes a strong second one once the team has learned how its tools and data behave.<\/span><\/p>\n<h2 id=\"h.63uo2cz9bi65\"><span>Why the top score is not always the right first project<\/span><\/h2>\n<p><span>A first project teaches you how your data, your tools and your staff react to automation, and that lesson is worth as much as the hours saved. SAP Signavio made this point in a 2015 post on <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/signavio.com\/post\/which-process-to-automate-first&amp;sa=D&amp;source=editors&amp;ust=1793628034803150&amp;usg=AOvVaw3-sbWcCvI4tjG7c8m1EeGr\" target=\"_blank\" rel=\"noopener\">choosing which process to automate first<\/a><\/span><span>, writing that &#8220;to the extent that the first process automation project is a learning exercise, it makes sense to choose a process where you can tolerate and easily correct mistakes.&#8221;<\/span><\/p>\n<p><span>Three other checks can move a candidate up or down. The first prioritisation criterion in the NSW guide is &#8220;their strategic alignment to your organisation&#8217;s goals,&#8221; so a process tied to something the business is trying to fix this year deserves a second look. It also lists stakeholder buy-in and executive sponsorship, because the people who do the work have to trust the result. And timing matters. A process that sits on a system you are about to replace, or one that only runs at year end, is hard to test and may need rebuilding soon after launch.<\/span><\/p>\n<h2 id=\"h.rhnozwew2gw0\"><span>Should the first process use AI?<\/span><\/h2>\n<p><span>Sometimes the most valuable candidate scores low on ease only because its input is free text, such as orders sent by email in the customer&#8217;s own words. An AI step can read that kind of input, which changes the ease score, but it also adds testing work and a new kind of mistake that a fixed rule would not make. The guide to <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/artinwebs.com\/blog\/intelligent-automation-dubai\/&amp;sa=D&amp;source=editors&amp;ust=1793628034805906&amp;usg=AOvVaw3htbi5EF1tmjzO-2k8wNmN\">intelligent automation in Dubai<\/a><\/span><span>&nbsp;covers when an AI step adds value and when plain automation is enough, and the advice holds outside the UAE too.<\/span><\/p>\n<p><span>When two candidates are close on value, choose the one that works without AI for the first project. If an AI agent is on your list, <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/artinwebs.com\/blog\/ai-agents-for-business-how-they-work\/&amp;sa=D&amp;source=editors&amp;ust=1793628034806812&amp;usg=AOvVaw0yK3W-mg8sg8H-CgKuurcT\">how AI agents for business work<\/a><\/span><span>&nbsp;explains which tasks suit an agent and when a workflow with a model inside is the better design.<\/span><\/p>\n<h2 id=\"h.cd2w9y2bwvob\"><span>Check the winner&#8217;s numbers before you commit<\/span><\/h2>\n<p><span>A score ranks candidates against each other. It does not tell you whether the winner pays for itself. Before you build, put real figures on it, including subscriptions, usage charges, maintenance and the staff time spent on failed runs. The guide on <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/artinwebs.com\/blog\/how-to-calculate-automation-roi\/&amp;sa=D&amp;source=editors&amp;ust=1793628034807975&amp;usg=AOvVaw2wYwsm1ik8Fp9BWc_reJFm\">how to calculate automation ROI<\/a><\/span><span>&nbsp;goes through each input and shows how a drop in volume can turn the result negative.<\/span><\/p>\n<p><span>For teams in the UAE, the <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/artinwebs.com\/tools\/ai-roi-calculator\/&amp;sa=D&amp;source=editors&amp;ust=1793628034808326&amp;usg=AOvVaw3Fobmt4-JffwSy8UrJ1tQd\">free AI ROI calculator<\/a><\/span><span>&nbsp;takes salary, visa, insurance and workspace costs in dirhams. Enter your own figures rather than relying on the example it shows when the page loads.<\/span><\/p>\n<h2 id=\"h.f9ygqsolyuru\"><span>The limits of any scoring method<\/span><\/h2>\n<p><span>Scores are only as good as the counts behind them. A guessed &#8220;about 20 minutes&#8221; turns into a precise-looking number on a sheet, and a ranking built on guesses can point at the wrong process with full confidence. Count runs and time for at least a normal week before you trust the result. Expect the first project to change your scores for the rest of the list, because you will learn which kinds of input and which systems were harder than they looked, and re-score the list after it goes live.<\/span><\/p>\n<h2 id=\"h.hfaxkjeleysj\"><span>What to do next<\/span><\/h2>\n<p><span>Once you have picked the process, the next step is to write down how it runs today, including every exception, before anything is built. That is the subject of the next guide in this series, on how to document a process before automating it.<\/span><\/p>\n<p><span>If an outside builder is likely, read <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/artinwebs.com\/blog\/hire-ai-automation-agency-dubai\/&amp;sa=D&amp;source=editors&amp;ust=1793628034810546&amp;usg=AOvVaw1XklnBz7fCjCsaW22aCdVS\">how to choose an AI automation agency in Dubai and what to ask<\/a><\/span><span>&nbsp;before your first call.<\/span><span>&nbsp;And if you want a second opinion on your short list, <\/span><span><a href=\"https:\/\/www.google.com\/url?q=https:\/\/artinwebs.com\/contact\/&amp;sa=D&amp;source=editors&amp;ust=1793628034810887&amp;usg=AOvVaw3YpuJGkBlw0EPe8DJ9ZtXS\">book a free 30-minute strategy call<\/a><\/span><span>&nbsp;and bring your scoring sheet and the counts behind it.<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Which process to automate first? Rule out poor fits, score each candidate on value and ease, weigh the cost of mistakes, then check the winner&#8217;s numbers.<\/p>\n","protected":false},"author":1,"featured_media":4451,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"rank_math_title":"","rank_math_description":"","rank_math_focus_keyword":"","rank_math_canonical_url":"","footnotes":""},"categories":[194],"tags":[195,203,176,196,169],"class_list":["post-4448","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-automation-planning","tag-automation-planning","tag-automation-prioritization","tag-business-automation","tag-process-automation","tag-small-business"],"acf":[],"_links":{"self":[{"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/posts\/4448","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/comments?post=4448"}],"version-history":[{"count":1,"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/posts\/4448\/revisions"}],"predecessor-version":[{"id":4452,"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/posts\/4448\/revisions\/4452"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/media\/4451"}],"wp:attachment":[{"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/media?parent=4448"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/categories?post=4448"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/artinwebs.com\/blog\/wp-json\/wp\/v2\/tags?post=4448"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}