--- title: "What should a business automate first: sales, collections, inventory, or management?" description: "A prioritization method for choosing the first automation by business impact, data readiness, risk, and the team's ability to handle exceptions." author: "Carlos García" published: 2026-09-06 updated: 2026-09-06 language: en human_url: "https://kiia.cloud/blog/what-to-automate-first-in-a-business/" --- > **In 60 seconds:** Do not start with the department that has the loudest demo. Score candidate processes on frequency, time consumed, error cost, effect on cash or customers, data quality, exception rate, and ownership. Pick one narrow slice with a clear trigger and outcome. Keep a manual fallback and review exceptions every week. The first project should teach the company how to automate, not force every process into AI. When a company says it wants to automate sales, collections, inventory, and management, the list is too broad to estimate or implement. Each area may contain dozens of tasks with different data, risk, and owners. The useful question is smaller: which recurring delay costs the company the most and is ready to be changed now? ## Build a list of processes, not departments "Automate sales" is not a process. "Create a follow-up task when a qualified opportunity has no activity for two business days" is one. So is "notify the collections owner when an undisputed invoice is overdue" or "show stock before a salesperson promises delivery." Ask the people doing the work where they retype information, wait for approval, search across systems, or discover an exception too late. Use real recent cases, including the messy ones. A clean workshop example hides the handoffs that make automation fail. For each candidate, write: - the event that starts it - the required data and source of truth - the rule or judgment applied - the action and expected outcome - the person responsible for exceptions - the events that stop or reverse the flow If the team cannot agree on these points, document the policy before adding software. ## Score value and readiness separately Give every candidate a score from one to five across two groups. Value includes frequency, manual time, error cost, delay to revenue, customer impact, and management visibility. Readiness includes source-data quality, rule clarity, API or connector availability, exception rate, security, and a named owner. A process with high value but low readiness needs preparation. A process with moderate value and high readiness can be the right first pilot because it produces evidence quickly. Avoid choosing a low-value process simply because the automation looks easy. The order-to-cash chain is a useful lens: stock visibility, quotation, order processing, invoicing, collections, payment, and reconciliation. Products such as [SAP Collections Management](https://help.sap.com/doc/0abb51690f15487a94e228274442cff4/2023.000/en-US/loioba7a71568b44215ae10000000a4450e5_PDF_EN.pdf) and [Oracle Collector Workspace](https://docs.oracle.com/en/cloud/saas/readiness/erp/26c/fins26c/26C-fin-wn-f49306.htm) show how rules and prioritization can organize collections. They do not prove that collections must be your first project. ## How the four common options compare Sales follow-up is attractive when opportunities are lost because no one owns the next action. It needs a clean CRM, contact rules, stop events, and message review. Our guide to [WhatsApp and Teams follow-up](/blog/automate-sales-follow-up-whatsapp-teams/) shows a narrow version. Collections can affect cash quickly, but disputed invoices and customer relationships make exceptions important. Start by prioritizing and preparing work for a person. Do not let a model invent a payment promise or decide a legal position. Inventory automation works well when people repeatedly check stock or copy movements. It fails when product codes, units, locations, or adjustments are inconsistent. Fix master data before automating promises to customers. A management assistant can reduce reporting work and expose bottlenecks. It depends on agreed definitions for revenue, overdue, active opportunity, and available stock. A chat interface does not resolve conflicting numbers. Start with a read-only briefing such as the [ERP-connected executive assistant](/blog/ai-executive-assistant-erp-sales-collections/). ## Choose the smallest valuable slice Suppose the team spends hours reviewing overdue invoices. The first slice might create a daily queue of invoices that are overdue, undisputed, above a threshold, and missing a scheduled follow-up. A collections owner checks the queue and acts. That is safer than autonomous collection messages and easier to measure. It also reveals whether due dates, disputes, owners, and contact history are reliable. Later iterations can draft messages or create approvals. The same pattern applies elsewhere. Prepare a qualified lead list before sending outreach. Show a stock exception before updating purchasing. Build a management briefing before allowing write actions. ## Measure the pilot against a baseline Before changing the process, record time per case, backlog, correction rate, missed deadlines, and one business outcome. During the pilot, track how much work enters the exception queue and how often people override the system. Set a review date. Keep the automation if it reduces meaningful work without raising errors or creating hidden risk. Narrow it if exceptions dominate. Stop it if nobody owns the queue. Do not promise a specific reduction in days sales outstanding, processing time, or headcount without evidence from your own operation. Vendor case studies are not your baseline. ## A practical first-week checklist Map ten recent cases, including failures. Confirm the system of record. Name the process owner and exception owner. Test duplicate events, missing fields, stale inventory, different currencies, local holidays, and manual holds. Add a log and a pause mechanism. Write the recovery steps before launch. Kiia helps teams turn a broad automation wish list into a sequence of bounded projects. The goal of the first project is a reliable result and a better map of the operation. That evidence tells you whether the next investment belongs in sales, collections, inventory, or management. ## Frequently asked questions ### Which department usually has the best first automation? There is no universal winner. Choose a frequent bottleneck with reliable data, a clear owner, visible impact, and a safe manual fallback. ### Should we begin with AI? Not necessarily. A rule, scheduled reminder, form, or integration may solve the first bottleneck more reliably and cheaply. ### How long should the first pilot run? Run enough real cases to include normal work and exceptions, then compare time, corrections, backlog, and business outcomes with the baseline.