AI-ready operations describe a firm whose core records (clients, projects, schedules, time, expenses and invoices) live in one connected system that people, integrations and AI assistants can all reach under clear permissions. For a professional services firm, that foundation is what lets teams automate freely without multiplying tools, credentials and competing versions of the truth.
Picture a Monday leadership meeting at a 120-person IT consultancy with offices in three cities. The delivery director has a utilisation forecast from an automation someone on her team wired up over a weekend. Finance has a different number, from a script that copies timesheets into a spreadsheet. The managing director wants an AI plan by the end of the quarter. Nobody in the room can say how many automations the firm runs, who owns them, or which number is right.
Scenes like that are what you get when AI use spreads faster than anyone measures it. The Thomson Reuters Institute’s 2026 AI in Professional Services Report, which surveyed more than 1,500 legal, tax, accounting, corporate and government professionals, found that organisation-wide AI use had almost doubled in a year, to 40% from 22%. Only 18% of respondents knew their organisation was tracking the return on those tools.
This guide covers how to let your teams keep building without losing one trusted set of numbers.
Key Takeaways
- AI-ready operations describe the state of your operational record and its controls, not the AI tools you’ve bought.
- Sprawl now comes in four forms (tools, automations, agents and unsanctioned AI), and cheap building makes each one grow faster.
- The working principle is to decentralise innovation and centralise operational truth, commercial logic and control.
- A six-stage maturity model and a 12-question scorecard show where your firm stands and what to do next.
What AI-Ready Operations Means
AI-ready operations are business operations in which the core operational record is connected, current, governed and reachable, so that people, integrations and AI assistants all work from the same facts under the same rules. The term describes your foundations, whether or not you’ve connected an AI tool yet.
- Connected: the client, the quote, the project, the people, the schedule, the time, the expenses and the invoice are linked records in one system, or in systems that sync automatically, not exports stitched together afterwards.
- Current: time and expenses are entered close to when the work happens, so an answer drawn from the record is still true when someone acts on it.
- Governed: every person and every connected tool works under permissions you can list, and you can say who is allowed to read, create, change or delete each kind of record.
- Reachable: the record is open to the tools your teams choose through documented routes such as APIs, webhooks and the Model Context Protocol (MCP), a standard way for AI assistants to connect to business systems. Our guide explains what an MCP server is.
Notice there’s no model, assistant or agent platform on that list. Readiness is a property of the record, which is why buying AI tools doesn’t create it.
Gartner predicted in February 2025 that through 2026 organisations will abandon 60% of AI projects unsupported by AI-ready data, a term it uses for data that is representative of the specific use case. The four conditions above are how an operational record gets there. In Salesforce’s 2026 Connectivity Benchmark Report, a survey of 1,050 IT leaders at enterprises with at least 1,000 employees, 96% said they face barriers to using data for AI, and 40% named outdated IT architecture caused by data silos and disconnected systems as a top blocker.
Why Easier Building Creates More Sprawl
Sprawl grows because building has become cheap while governing hasn’t. AI coding tools and no-code automation let any team create a workflow, an integration or an agent in an afternoon, and each one tends to arrive with its own copy of the data, its own credentials and no named owner.
The by-products are familiar. You find duplicated workflows, scattered credentials, unclear ownership, conflicting data, abandoned automations, inconsistent permissions and a growing governance load on your IT and systems leads. The easier it becomes for every department to build, the more important a governed system of record becomes. A system of record is the one authoritative home for operational data.
In an October 2025 Zapier survey of 550 US executives at companies with 1,000 or more employees, 28% said their organisation uses more than 10 different AI apps, and only 35% said those tools go through proper approval channels. Salesforce’s 2026 Connectivity Benchmark Report, covering enterprises of the same size, found that only 27% of enterprise applications are integrated and that half of AI agents operate in isolated silos.
A 100-person consultancy is a different scale, but the same cheap tools are in your teams’ hands. Four kinds of sprawl are worth telling apart.
Kind of Sprawl | What It Is | What It Looks Like in a Services Firm |
|---|---|---|
Tool sprawl | Overlapping software piling up. TechTarget defines it as “the accumulation of many IT tools by an organization, leading to inefficiency and data siloing.” Counted in subscriptions, it’s often called SaaS sprawl. | Projects in one tool, time in another, resourcing in a spreadsheet and invoicing in the accounting system. |
Automation sprawl | Workflows, scripts and integrations built team by team with no shared register. | Three separate jobs that each copy timesheets somewhere, built by people who have since changed roles. |
Agent sprawl | AI agents and assistants multiplying faster than anyone lists them. | A proposal agent, a status-report agent and a billing helper, each with its own access and its own idea of project status. |
Shadow AI | IBM defines it as “the unsanctioned use of any artificial intelligence (AI) tool or application” without IT approval. Shadow IT is the wider category covering any unapproved application. | Client details pasted into a personal AI account to draft a status update. |
What Sprawl Costs a Services Firm
Sprawl costs a services firm in three places. Billable work fails to reach an invoice, capacity becomes hard to see, and senior people spend hours reconciling figures that should have matched.
When hours are logged in one tool and rates live in another, some billable time never makes it across. When the schedule isn’t linked to the projects it serves, nobody can see who has capacity next month. And when two automations each calculate utilisation from their own copy of the timesheets, leadership ends up debating which number to believe.
In Zapier’s October 2025 survey of US enterprise executives, 76% of enterprises had experienced at least one negative outcome from disconnected AI, 36% of leaders said AI sprawl was increasing security and privacy risks, and 30% said they were wasting money on redundant AI software.
We couldn’t find a reliable published figure for mid-sized services firms, so count the hours your finance and delivery leads spend reconciling reports each month.
Signs Your Firm Already Has Sprawl
- Delivery and finance report different revenue figures for the same project.
- Nobody can produce a list of the automations and AI tools that touch client or billing data.
- API keys or logins are shared in chat threads or written into scripts.
- Month-end billing depends on a spreadsheet only one person understands.
- An AI assistant answered a project question confidently and wrongly, because it could see only part of the record.
Decentralise Innovation, Centralise Operational Truth
The way to control sprawl without slowing teams down is to decentralise innovation and centralise operational truth, commercial logic and control. Teams keep building while the record and the rules stay in one place.
Blocking isn’t the fix. Gartner analyst Max Goss has said that “many organizations resort to blocking or restricting the use of AI agents, but this is not a long-term solution.”
The centre is your system of record, the system holding the authoritative version of clients, projects, rates, time and invoices, and the billing rules applied to them. A system of intelligence analyses or acts on those records but shouldn’t own them. In professional services, the category built to be that centre is professional services automation (PSA), which connects the chain from quote to project to time to invoice.
Depending on one platform is a fair worry, so the centre has to be open, with a flexible edge of assistants and team-built workflows around it. Our guide to building, buying or extending software covers that split.
The AI-Ready Operations Maturity Model
The AI-Ready Operations Maturity Model describes six stages a services firm moves through, from manual, spreadsheet-led work through to an emerging sixth stage of AI-enabled operations across connected processes. Each stage sets a ceiling on what AI can safely do, because an assistant is only as reliable as the record it reads.
Stage | What It Looks Like | What AI Can Safely Do | Next Move |
|---|---|---|---|
1. Manual and spreadsheet-led | Information is scattered across spreadsheets and inboxes, and reporting is assembled by hand. | Help individuals draft and summarise. Anything it says about the business needs checking. | Get projects, time and billing into systems before automating anything. |
2. Multiple disconnected tools | Projects, time, expenses and billing sit in separate systems, with manual reconciliation and unclear ownership. | Answer questions inside one tool at a time. Answers that span tools conflict. | Choose the system of record and consolidate the core chain into it. |
3. Central operational platform | Core operational and commercial records are connected in one system of record. | Answer questions across projects, people, time and billing from one source. | Open governed access through APIs and webhooks, and start a register of automations. |
4. Managed integrations and automation | APIs and workflows remove repetitive work, and each integration has an owner and documentation. | Feed reporting and analytics reliably, and support read-only assistants. | Set permissions for AI access before connecting assistants that can act. |
5. Governed AI assistants | Assistants reach approved operational data through connections such as MCP, under controlled permissions. | Retrieve, summarise and prepare work, and take defined actions within each user’s permissions. | Extend action rights area by area as trust builds. |
6. Cross-functional AI-enabled operations | Emerging direction. AI and automation run across connected processes while data and commercial controls stay central. | Emerging direction. Agents reason across staffing, margin and billing, with people keeping commercial judgement. | Keep the core stable and keep building at the edge. |
Stages 1 to 5 are achievable today. Stage 6 is a direction, not something you can buy.
Take the 120-person IT consultancy from the opening, with three offices, mixed time-and-materials and fixed-fee work, and automations built team by team on separate project, time and billing tools. That firm is at Stage 2, whatever its AI ambitions, and its numbers disagree because its automations have run ahead of the record they depend on. Its next move isn’t an agent platform. It’s consolidating delivery, time and billing into one system of record across all three offices.
Start at the stage you’re actually at. Few firms are as far along as the headlines suggest. In SPI Research’s 2025 study of AI in professional services, which covered 146 organisations, only 40% of consultants and employees had access to AI, and about one in ten firms were attempting transformational AI initiatives.
How to Automate Without Creating Tool Sprawl
To automate without creating tool sprawl, fix the operational record first, give every fact one home, register every automation with a named owner, connect through governed access instead of copied credentials, and review what you’ve built on a schedule. Seven rules cover it.
- Fix the record before you automate it. An automation built on incomplete timesheets or out-of-date rates produces wrong answers faster. Get time and expenses entered and approved reliably first, in a system people will actually use.
- Give every fact one home. Decide which system is the source of truth for each kind of record, such as clients, rates, hours and invoices. Automations read from that home and write back to it without keeping a private copy.
- Register every automation with a named owner. Keep one list showing what each automation does, what it reads and writes, whose access it uses, who owns it and when it was last reviewed. Gartner’s six steps for managing AI agent sprawl include a centralised agent inventory, and Gartner reports that only 13% of organisations think they have the right agent governance in place.
- Connect through governed access, not copied credentials. Use documented routes instead of shared logins. APIs read and write records, webhooks notify other systems when a record changes, and MCP connections serve AI assistants. Make every connection that signs in do so as an identifiable user or service. Our comparison of MCP, APIs and webhooks covers which route suits which job.
- Apply least privilege to every connection. Least privilege means each person, integration and assistant gets the minimum access the job needs. Start assistants read-only, widen their rights area by area, and decide in advance which actions assistants may never take. Our guide to AI governance at work covers the control layers.
- Review and retire on a schedule. Each quarter, switch off automations nobody uses and remove access when owners change roles.
- Judge automations by operational outcomes. Measure missing timesheets, time to invoice and hours spent reconciling, not the number of agents deployed.
The same seven rules prevent automation sprawl and agent sprawl too, and none of them needs a central team approving every idea.
The AI-Ready Operations Scorecard
The AI-Ready Operations Scorecard is a 12-question check you can run with your delivery, finance and IT leads in a single meeting. It works as a quick AI readiness assessment for a services firm. Score one point for each honest yes.
Connected Record
- Do clients, projects, schedules, time, expenses and invoices live in one system, or in systems that sync automatically?
- Can you trace an invoice line back to the time entry and project behind it without opening a spreadsheet?
- Do delivery and finance report utilisation and revenue from the same source?
Current Data
- Is most time logged in the week it was worked?
- Are timesheets and expenses approved before they’re billed or reported?
- Are closed periods locked so reported figures don’t change afterwards?
Ownership and Inventory
- Can you list every automation, integration and AI tool that touches client or billing data?
- Does each one have a named owner who still holds that role?
- Do you review the list at least quarterly and switch off what’s no longer used?
Governed Access
- Do connected tools reach the record through documented APIs or connectors, each signing in as an identifiable user or service?
- Can you restrict what connected tools may read, create, update and delete, by area of the system?
- Have you decided which actions assistants may never take, and written it down?
Score | Likely Stage | What to Do Next |
|---|---|---|
0–3 | Stages 1–2 | Consolidate the core record before automating anything else. |
4–6 | Stage 3 | Open governed access and start the automation register. |
7–9 | Stage 4 | Set AI permissions, then connect read-only assistants. |
10–12 | Stage 5 | Extend what assistants may do, one area at a time. |
Treat the result as a conversation starter, and look at which of the four groups scored lowest. If you scored zero or one on Connected Record, treat your firm as Stage 2 whatever the total.
Maturity tracks with results on SPI Research’s own, broader scale. Its 2026 Professional Services Maturity Benchmark, covering 509 services organisations, reports that Level 5 firms achieve 42% more billable utilisation than Level 2 firms.
Where a Connected PSA Platform Fits
A connected PSA platform sits at the centre of the approach described in this guide, as the system of record that people, automations and assistants read from and write to under permissions. It doesn’t make a firm AI-ready on its own, but it provides the part that is hardest to build yourself, one governed place for the operational chain.
Avaza is a connected PSA platform for services firms of 20 to 300+ people running multiple teams and billing models. It holds the chain from customer and quote through project, people, schedule, time and expenses to invoice and reporting in one system, and connects to accounting tools such as Xero and QuickBooks through its integrations. Our PSA buyer’s guide covers how to evaluate the category.
INFORM DataLab, a German data and IT services firm of more than 50 people, is a working example of the reachable condition, using Avaza’s APIs and webhooks to feed its own Snowflake warehouse from the operational record it keeps in Avaza.
For firms exploring AI assistants or internal automation, Avaza also provides MCP access across connected project-to-cash workflows, covering projects, tasks, schedules, timesheets, expenses, estimates, invoices, approvals and payments. Any user can authorise a compatible assistant under their own account, the connection can’t exceed that user’s existing permissions, and administrators can narrow it further through API Permissions, by module and by action. That gives teams a governed way to extend their operational system.
There is no general Avaza review queue where every API or MCP action waits for approval. However, existing timesheet and expense approval workflows still apply to records submitted through an assistant, and organisations can make those approvals mandatory. AI submissions do not bypass those workflows. Firms requiring human review can restrict Create/Update/Delete access for the relevant modules, account-wide or per user, through API Permissions. Through Avaza’s MCP, a delete is a two-step action in which the assistant previews the deletion and then requires a separate confirmation. API Permission changes are logged and exportable, and connected tools operate under identifiable user permissions.
Today, a compatible assistant connected to Avaza can check project status, find timesheets, review billing, approve or reject timesheets and submit expenses for approval, all within the user’s permissions. As governed agents mature, the opportunity is for agents that reason across staffing, margin and billing together. That is an emerging direction, not a shipped capability. AI access isn’t yet the reason most firms choose a PSA platform, but it’s an increasingly relevant evaluation criterion. Disconnected project, capacity, time and billing data remains a main reason.
Frequently Asked Questions
What Are AI-Ready Operations?
AI-ready operations are business operations whose core records are connected, current, governed and reachable, so people, integrations and AI assistants work from the same facts under the same rules. In a professional services firm, that means clients, projects, schedules, time, expenses and invoices held in one system of record with clear permissions.
What Is the Difference Between AI Readiness and AI-Ready Operations?
AI readiness is the broader term. It usually covers a whole organisation’s preparedness for AI, including strategy, skills, culture and data. AI-ready operations is narrower. It describes whether the operational record and its controls can support automation and AI assistants reliably. A firm can have an AI strategy without having AI-ready operations.
What Is the Difference Between Shadow IT and Shadow AI?
Shadow IT is any application or service used without IT approval. Shadow AI is the AI-specific version. IBM defines it as the unsanctioned use of any AI tool or application. In a services firm, shadow AI often means client or project data entered into personal AI accounts.
Do You Need Perfect Data Before Using AI in a Services Firm?
No, not everywhere at once. Gartner analyst Roxane Edjlali has said AI-ready data “must be representative of the use case,” so readiness is established one use case at a time. If timesheets are complete and current, an assistant can already help with questions about time and billing while other records are still being tidied. Start where the record is strongest.
How Should a Multi-Office Firm Approach AI-Ready Operations?
A multi-office firm should agree one system of record and one automation register across every office before connecting assistants. Offices tend to build their own workflows and copy the same record. Standardise how time, expenses and their approvals work everywhere, then let each office build its own workflows and assistants on top, under the same permissions.
Should You Stop Teams From Building Their Own Automations?
No. Gartner analyst Max Goss warns that employees who can’t work in sanctioned tools are likely to go around the controls and turn to shadow AI. Let teams build on a shared record, with every automation registered to a named owner and every connection running under identifiable, limited permissions. Control the data and the access, not the ideas.
Start With the Record
AI-ready operations aren’t something you buy with an assistant. They come from one connected, current, governed and reachable operational record and the seven rules that keep automation owned and governed. Run the scorecard, find your stage and take the next move for it.
If you’d like help with that, book a call and we can map where your firm stands and how Avaza would fit your delivery, time and billing workflows. You can also start a free 14-day trial, no credit card required, and test MCP straight away.
