How to Run a Tech Stack Audit: Map Your Systems, Processes, Apps, Integrations, People, AI and Automation

A tech stack audit is a structured review of your systems, processes, apps and integrations, and of the automation and AI that run across them. Done properly, it turns a sprawl of subscriptions into a deliberate, measurable operating system for the business.

Most SMEs accumulate software reactively. A CRM is bought for sales, an email tool for marketing, a scheduling app for a new hire, a Zapier account to glue them together, and now a handful of AI subscriptions on top. Three years later, nobody can say which system holds the true customer record, which automations are still running, or why the same lead receives two welcome emails.

The case for auditing is stronger in 2026 than ever. All-in-one platforms such as HubSpot and HighLevel now bundle CRM, email, booking, payments and AI agents, while Make, Zapier and n8n have become AI orchestration layers. The question is no longer "which tool should we add?" but "which tools can we retire, and what should automation and AI do with the rest?"

The six layers of a complete audit

A useful audit examines every layer of how the business runs, not just the software invoice:

• Systems: the core platforms and the data each one owns.

• Processes: how work actually moves from enquiry to cash and beyond.

• Apps: what each tool does, and how much of it you use.

• Integrations: how data moves between apps, and how reliably.

• Automation: the workflows already running, and the manual work still waiting to be automated.

• AI: the AI tools and features in use, how safely they handle data, and where AI could add real value.

Across all six layers, the audit also looks for overlap, so you can consolidate your tech stack and stop paying for the same capability twice.

The questions a good audit answers

By the end, you should have an evidenced answer to each of the following:

1. Which applications do we use, who owns them and what do they cost?

2. What functionality does each application provide, and how much of it do we use?

3. How does information move between systems, and which system owns each record?

4. Which automations are running, are they reliable, and what should be automated next?

5. Where is AI being used, is it safe, and where could it add the most value?

6. Which tools and features overlap, and what should be consolidated?

7. What should we improve first, and how will we measure success?

The ten-step method below is the one we use at OpScaling for our Automation and AI Audits, built around our audit workbook: a systems inventory, an app integration register and departmental process worksheets.

Get a Free Audit Template

Open this Google sheet to get a free a Free Business Audit Template

Step 1: Set the scope, objectives and baseline

Start with the outcomes you want to improve, because "review our software" is too broad to guide useful decisions. Strong objectives include reducing repeated data entry, improving lead response times, shortening onboarding, making reporting reliable, cutting subscription costs or preparing the business for AI.

Choose a process boundary. An enquiry-to-payment audit covers marketing, sales, proposals, delivery and finance; an employee onboarding audit involves a different set of systems and owners.

Record the baseline before changing anything

Capture measurable starting points so every improvement can be proven later:

• Processing time for key workflows, such as enquiry to first response.

• The number of manual steps and handovers in each process.

• Duplicate or incomplete records in the CRM.

• The number of active automations, and how many have failed recently.

• Total software spend, including add-ons, usage charges and AI subscriptions.

Where figures are estimates, label them as such.

Speak to the people doing the work

Interview the staff who run each process, not only department managers. A documented process may claim that customer data synchronises automatically, while the team quietly maintains a spreadsheet because the integration misses important fields. Ask the same people which AI tools they use day to day; the answer is often more than leadership expects.

Step 2: Audit your systems and data ownership

Your systems are the core platforms the business depends on: typically the CRM, accounting platform, project management tool, website and communication suite. This step establishes what they are, who owns them and which data each one is responsible for.

List every system by business function

Organise the inventory by function rather than by vendor, because grouping by function exposes gaps and duplicates that a subscription list hides. We audit across eleven areas: Sales, Marketing, Website and Funnels, Membership, Service Delivery, Operations, Artificial Intelligence, Reporting, Project Management, Finance and HR.

End each area with a line for offline processes, such as paper forms, emailed spreadsheets and manual invoice runs. These are often the highest-value automation opportunities.

Find the shadow IT

Cross-check the list against card statements, the accounting system and single sign-on logins. Teams routinely subscribe to tools, including AI assistants, on personal or departmental cards, and these rarely appear in any register.

Decide which system owns each type of data

You do not need every record in one application, but you do need one source of truth for each type of data:

• Contacts, deals and sales ownership: usually the CRM.

• Invoices, payments and balances: the accounting platform, with payment status shared back to the CRM.

• Job progress and deadlines: the project management platform.

• Signed agreements: the contract or document system.

• Reporting measures: a governed reporting model, so every dashboard uses the same definitions.

Where two systems both claim ownership of the same data, you have found a sync conflict, and usually the root cause of duplicate records and inconsistent reporting. Set ownership at field level where needed: sales may own the primary contact, while finance owns the billing address.

Step 3: Map your business processes

An inventory explains what you own; a process map explains how the business uses it. Choose a real example, such as a recent enquiry that became a customer, and follow it end to end, including emails, approvals, spreadsheets and phone calls alongside the automated steps.

Use the six-question framework

The departmental worksheets in our audit workbook document every step with six questions:

• How: the steps involved, and whether each is manual or automated.

• What: the data required, such as name, email, deal value or service type.

• Why: the purpose of the step and the benefit of automating it.

• Who: the person, team, customer or system responsible.

• When: the event, timing or dependency that starts the step.

• Where: the application in which the work happens.

Document the current process before designing the improved one; otherwise it becomes difficult to separate existing capability from proposed change.

Example: an enquiry-to-onboarding process

This illustrative example shows how an apparently simple sales process can hide several manual handovers:

1. Enquiry received (website form): the customer submits their details. Does every submission reliably reach the CRM?

2. Lead recorded (CRM): a contact and opportunity are created. Are existing records checked first?

3. Qualification (CRM and email): sales reviews the requirements. Are the criteria and ownership clear?

4. Proposal prepared (proposal app): staff copy customer and pricing data. Which fields could populate automatically?

5. Agreement signed (e-signature app): sales receives a notification. Does the signed status reach the right opportunity?

6. Job created (project management): operations creates the project by hand. Are scope, files and deadlines transferred?

7. Invoice prepared (accounting platform): finance re-enters customer details. Is the customer matched to an existing account?

Every stage where a person copies data from one system to another is a candidate for automation, particularly where responsibility moves between departments. Mark each one on the map; these become the shortlist for Step 6.

Step 4: Audit your apps and the functionality you use

Most businesses use a fraction of what they pay for. This step records what every app actually does, how deeply it is used, and which capabilities are sitting unused inside platforms you already own.

What to record for each app

Capture the same attributes for every application so the results can be compared and scored:

• Application and purpose: the product, plan and business problem it solves.

• Department and owner: who uses it, who administers it and who approves changes.

• Usage level: Basic, Moderate or Advanced, plus active users and frequency.

• Cost and renewal: seats, add-ons, usage charges and renewal date.

• Data held: customer, financial, employee or operational information.

• Dependencies: the integrations, automations and reports that rely on it.

• Limitations: missing features, workarounds and frustrations.

• Business criticality: what stops working if the app becomes unavailable.

Record each app once, then associate it with every function it serves, so a CRM is not counted separately under sales, marketing and service.

Classify every feature into one of four states

App categories hide significant overlap: a CRM may also handle booking, forms, proposals, chat, client portals and AI agents. For each relevant feature, record its state:

• Used: configured and supporting a live process.

• Available: included in your current plan but unused.

• Upgrade required: available only on another plan or add-on.

• Not suitable: absent, or unable to meet the requirement.

The "Available" list is where the quickest wins usually hide, both for consolidation and for automation. Verify important capabilities in your actual account and the vendor's current documentation, and note the plan and date checked; a feature on a vendor's website is not proof that your subscription includes it.

Step 5: Audit your integrations

"These apps integrate" is only the start of the assessment. You need to know what each connection transfers, in which direction, how often, under what conditions, and what happens when it fails.

Build an integration register

The App Integration worksheet records one entry per connection: the department, source app, destination app, trigger event, actions, step type and whether the trigger is instant or scheduled. For example:

• Finance, Stripe to Xero: when a new Stripe payment is received (instant), create the invoice in Xero and allocate the payment.

• Sales, Calendly to CRM: when a meeting is booked (instant), record the booking on the company record.

• Marketing, website form to CRM: when a form is submitted (instant), create or update the contact and start the nurture sequence.

Extend each entry with the fields transferred, the account owner, the date of the last successful test and the failure response.

Classify how each connection is built

The build method determines an integration's fragility, cost and who can maintain it:

• Native integration: a built-in connection offered by the platform. Check the supported records, fields, sync direction and plan requirements.

• Marketplace connector: an app installed from a vendor or third-party marketplace. Check the publisher, support, permissions and extra charges.

• Automation platform: a workflow in Zapier, Make or n8n. Check the triggers and actions available, task or operation costs, and error handling.

• Custom API or webhook: code or event notifications sent directly between apps. Check authentication, rate limits, duplicate delivery, retries and who maintains it.

• Embedded interface: one app displayed inside another. Check whether data is actually exchanged or only displayed.

• Manual transfer: CSV exports, copy and paste or re-keying. These are prime automation candidates.

Match records on stable identifiers

Email addresses help identify contacts, but they change, get shared and can relate to several opportunities. Store each system's record ID against the others, and agree which system wins when a two-way sync conflicts. This matters most in finance automation, where a technically successful transfer must still link the correct customer, invoice and payment.

Step 6: Audit your automation and get more from it

Automation is where most businesses either gain or lose the most value from their stack. Many SMEs already run dozens of workflows across Zapier, Make, n8n and their CRM, often built by different people at different times, with nobody responsible for the whole picture.

This step has two goals: make the automation you already have reliable and efficient, then find the highest-value work still being done by hand.

Inventory every automation you already run

List every workflow across every platform, including native CRM workflows, Zapier Zaps, Make scenarios, n8n workflows and app-level rules. For each one, record:

• Owner: who built it, and who maintains it now.

• Purpose: the business outcome it supports.

• Trigger and steps: what starts it and what it does.

• Volume and cost: monthly runs, and the tasks or operations it consumes.

• Health: the last error, the error rate and whether anyone is alerted when it fails.

• Still needed?: automations built for old processes, departed staff or retired apps often keep running unnoticed.

Break each workflow into its components

Automation becomes far easier to assess once each workflow is described in the same language the platforms use:

• Trigger: the event that starts the workflow, such as "deal moved to Won".

• Search: a lookup that finds an existing record before acting, preventing duplicates.

• Conditions: the rules that decide whether it continues, or which route it takes.

• Actions: what changes as a result, such as "create project" or "send email".

• Timing: instant (event-driven) or scheduled, with an agreed maximum delay.

• Stop conditions: what ends it, so a cancelled booking stops its reminders and a signed proposal stops chaser emails.

As an example, a well-specified onboarding workflow is triggered when a proposal is signed, searches for the CRM opportunity using the stored proposal ID, checks that all signatures are complete and onboarding has not already started, then updates the deal to Won, creates the project from a template, assigns the delivery owner, drafts the invoice, sends the welcome email and writes the project and invoice IDs back to the CRM. If any step fails, it alerts the account owner.

Find duplicate and conflicting automations

Overlapping automations are as costly as overlapping apps. A booking tool, the CRM and a Zap might each send a confirmation for the same event, or two systems might update each other in an endless loop. Assign one owner to every outcome and switch off the rest.

Test reliability, not just the happy path

A workflow that works once in a demonstration may still fail in daily operation. Test each critical automation against missing fields, duplicate submissions, expired authentication, an unavailable destination app and partial completion, such as a project created but the invoice step failed. A retry must resume safely without creating a second project or invoice.

Verify results in the destination system rather than trusting a green tick; the wrong contact or an empty field can pass through a technically successful run. Mark untested automations as unverified. Our Zapier tips and Make.com tips cover error handling in more detail.

Get more from the automation you have

Once the existing estate is stable, these actions typically release the most value:

• Move work to native features where your CRM now does the job, cutting middleware task costs and failure points.

• Merge fragmented workflows into fewer, well-documented scenarios with clear owners.

• Add monitoring and alerts so failures are caught within hours, not weeks.

• Document every workflow in your SOPs, so the business does not depend on whoever built it.

• Know the limits: some processes outgrow no-code tools, so understand the limitations of no-code solutions before building something critical.

Find the next automation opportunities

Return to the manual steps flagged in Step 3 and score each one on frequency, staff effort, business impact, rule clarity and data readiness. The best candidates are frequent, predictable tasks where delays or mistakes carry a real cost, such as lead response, follow-up, booking, invoicing and onboarding.

For a quick capacity estimate, multiply monthly task volume by minutes saved per task and divide by 60. Saving four minutes on 300 tasks a month releases 20 hours, before monitoring and exceptions; this is an illustration, not a guaranteed saving. See our SMB automation services for how we prioritise and build these workflows.

Step 7: Audit your AI and get more from it

AI has entered most businesses faster than any previous technology, usually through individual subscriptions rather than a plan. An AI audit establishes what is in use, whether it is safe, where it overlaps, and where it could deliver far more value than it does today.

Inventory every AI tool and feature

AI now lives in three places, and the audit should cover all of them:

• Standalone assistants: ChatGPT, Claude, Gemini or Copilot seats, whether paid by the business or by individuals.

• Embedded AI features: AI built into your CRM, helpdesk, email, design and meeting tools, which is often switched on by default.

• AI steps in automations: LLM calls inside Zapier, Make or n8n workflows, and any custom AI agents or chatbots.

For each, record who uses it, what it costs, what data it receives, what it produces, whether it can change business records and who reviews its output.

Check data protection and governance

AI raises questions that ordinary software does not. Confirm which customer, employee or financial data is sent to which model, whether business plans with appropriate data terms are in place, and whether staff are pasting sensitive information into personal accounts. Agree a short acceptable-use policy and assess each use case against UK GDPR.

Remove AI overlap

AI sprawl is the newest form of tech stack overlap. A typical SME may pay for several chat assistants, an AI writing tool, an AI meeting note-taker and the AI add-ons in its CRM, all doing similar work. Standardise on one approved assistant plan and use the AI built into your core platforms wherever it meets the requirement.

Match AI to the right tasks

The key distinction is between fixed rules and interpretation. Creating a project after a signed proposal follows a rule, so it suits conventional automation. Tasks that involve reading, judging or writing suit AI, as our guide to automation versus AI explains. High-value AI use cases for SMEs include:

• Summarising sales and client calls, and logging next steps to the CRM.

• Classifying and routing inbound enquiries and support requests.

• Extracting data from invoices, contracts and emails into structured records, by combining natural language with structured data.

• Drafting proposals, replies and follow-ups for human review.

• Answering staff and customer questions from your own knowledge base, using retrieval-augmented generation (RAG) or by turning SOPs into AI assistants.

Keep a human in the loop and measure results

For each AI step, define the expected output, the review required, the failure route and the cost per run. Test it on representative examples before its output drives important actions, and send uncertain cases to a person. Our guide to human-in-the-loop AI systems covers how to design this. Measure accuracy and correction rates as well as time saved, because an AI step that needs frequent fixing can add work.

Prepare for AI agents

The next stage is AI agents that take bounded actions across your systems, connected through Make, Zapier, n8n and the Model Context Protocol (MCP). Agents depend on the same foundations as every other layer of this audit. Your business is ready for them when:

• Each type of data has one clear source of truth (Step 2).

• Processes are documented, with rules and exceptions defined (Step 3).

• Integrations are reliable and monitored (Steps 5 and 6).

• AI use is governed, with approved tools and data rules (this step).

Our AI consultants work across OpenAI, Anthropic and Gemini models to design AI workflows that are useful, safe and measurable. For a primer, read large language models for SMBs.

Step 8: Consolidate your tech stack and remove overlap

With systems, apps, integrations, automation and AI all mapped, overlap becomes visible. Consolidation means removing duplicate subscriptions, moving work into platforms you already own, and standardising teams on shared tools, so you pay for each capability once and maintain fewer connections.

Separate functional overlap from operational duplication

Functional overlap means several apps can perform the same task; operational duplication means people perform the same work repeatedly. Two apps that both send email are not automatically redundant, but staff updating the same customer details in three places is a clear signal to act.

Common overlap hotspots

In our audits, the same categories recur across almost every business:

• Live chat: Intercom, Drift or LiveChat, when the CRM chat widget in HubSpot or HighLevel may already cover it.

• Scheduling: Calendly or Acuity, alongside CRM calendars and booking pages.

• Email marketing: Mailchimp or ActiveCampaign, alongside CRM marketing email.

• Forms and surveys: Typeform, Google Forms or SurveyMonkey, alongside CRM or website forms.

• Proposals and e-signature: PandaDoc, Proposify or DocuSign, alongside CRM quotes and documents.

• Landing pages: Unbounce, ClickFunnels or Leadpages, alongside CRM funnel builders.

• Project and task management: ClickUp, Asana and Trello used side by side.

• Knowledge and SOPs: Notion, Confluence and Google Docs, where one wiki would do.

• AI assistants and add-ons: several chat subscriptions plus AI features in every app.

• Automation platforms: Zapier and Make both in use, with no clear rule for which handles what.

Decide what to keep, migrate or retire

Assess each overlapping tool on two dimensions: how far its features overlap with your core platform, and how critical it is to revenue or delivery. Each tool then falls into one of four decisions:

• Low criticality, high overlap: retire now. The native feature is already paid for, so switch over and cancel at renewal.

• High criticality, high overlap: plan a phased migration. Consolidate onto the core platform with data mapping and a period of parallel running.

• High criticality, low overlap: keep and integrate. A specialist tool the core platform cannot match; harden the integration instead.

• Low criticality, low overlap: review usage. Retire it if adoption stays Basic by the next renewal date.

Watch the software market

Three market forces should inform every consolidation decision. All-in-one platforms keep expanding into email, booking, payments, memberships and reporting. AI is being built into the core of those platforms, reducing the need for separate point solutions. And integration platforms are becoming AI orchestration layers that connect tools and AI agents.

Vendor acquisitions, product mergers and retirements can also change pricing, support and integration options. Base decisions on confirmed, tested capabilities, and treat roadmap promises as assumptions until they ship.

Weigh the dependency risk

Consolidation concentrates dependency, so consider the effect of an outage, a price rise or a future migration, and confirm you can export your data. Keeping an effective specialist tool and integrating it properly is sometimes the right answer. Our platform consulting services cover both app consolidation and platform selection.

Step 9: Compare the full cost and payback

Subscription fees are only part of the calculation. Include paid seats, plan upgrades, AI subscriptions and usage charges, Zapier tasks or Make operations, integration maintenance, manual administration, training and support.

For any change, also include migration, data cleansing, rebuilding automations and any period when both systems must run in parallel. Align cancellations with renewal dates to avoid paying twice.

A worked payback example

The following hypothetical example shows the calculation:

• Subscriptions removed save £180 per month.

• Additional platform and integration costs add £60 per month.

• The net recurring saving is therefore £120 per month.

• One-off migration and implementation costs £1,200.

• The simple cash payback is 10 months (£1,200 ÷ £120).

This assumes savings begin immediately and costs stay unchanged; cancellation terms and parallel running could extend it.

Report time savings separately

Time released by automation and AI creates capacity, but it does not automatically reduce payroll. State how the capacity will be used, such as faster lead follow-up, more billable delivery or reduced overtime, so the benefit is real rather than theoretical.

Step 10: Prioritise and build your roadmap

Assign every app, integration, automation and AI tool a clear decision: keep, optimise, integrate, automate, consolidate, replace or retire. Then sequence the work by business impact, urgency, effort, risk and dependencies.

1. Immediate fixes: repair failed integrations, switch off duplicate automations, stop duplicate messages and assign missing owners. Success means correct outcomes shown with test records.

2. Quick automation wins: automate the highest-scoring manual steps from Step 6 using features you already pay for. Success means shorter processing times and fewer manual steps.

3. AI pilots: test one or two governed AI use cases with human review. Success means measured accuracy, acceptable correction rates and real time saved.

4. Consolidation pilots: test replacement functionality with a small user group. Success means the required features and exception cases pass agreed checks.

5. Structural changes: migrate a core platform or redesign the data model. Success means reconciled records, working integrations and team acceptance.

Manage the change safely

Give each change an owner, acceptance criteria and a rollback plan. Map dependencies before retiring an app, because it may power a form, report or automation that is easy to overlook.

Keep the old system available until data, workflows and access have been verified in the new one. Train users on the revised process, update your SOPs and assign ongoing responsibility for monitoring. For reporting on progress, a Looker Studio dashboard keeps the baseline and results in one place.

What your completed audit should contain

The finished audit should give leadership a set of evidence-based outputs they can act on immediately:

1. Systems inventory: every app by function, with owner, plan, usage level, cost and renewal date.

2. Capability matrix: features classified as used, available, upgrade required or not suitable.

3. Process maps: current and proposed workflows, with manual steps flagged.

4. Integration register: every connection with its trigger, actions, timing, method and test status.

5. Automation register: every workflow with its owner, volume, cost, health and next opportunities.

6. AI assessment: tools in use, data and governance findings, overlap and recommended use cases.

7. Data ownership rules: the source of truth for each record type and key field.

8. Consolidation plan: keep, migrate or retire decisions for every overlapping tool.

9. Prioritised roadmap: sequenced actions with cost, payback and success measures.

Each recommendation should state the problem, the proposed change, the expected benefit, the cost and the evidence needed to confirm it works.

Make it a recurring discipline

Treat the audit as a living register, not a one-off project. Review it quarterly, reassess before major renewals or migrations, and check every new tool and AI subscription against it first. As AI capabilities keep moving into core platforms, the businesses that review their stack regularly will run leaner, report more accurately and automate faster than those that do not.

Book a tech stack audit with OpScaling

OpScaling runs Automation and AI Audits using the framework above, covering systems, processes, apps, integrations, automation and AI across eleven business functions. We then design and build the resulting workflows in CRM. PMS using tools such as Make, Zapier and n8n, combining quick fixes with a longer-term roadmap for automation, AI and growth.

See the results we have delivered in our case studies, or book your free consultation to get a clear view of your tech stack.