Data Preparation & Advisory
Insightful AI audits, cleans, and structures organisational data so your AI tools produce outputs you can trust. We work with what you already have, across whatever systems you use, and we hand you documented, governed, AI-ready datasets with a clear maintenance plan. Fixed price from £3,000 ex VAT, scoped following a discovery workshop.
The data readiness journey
Scattered sources
Spreadsheets, inboxes and systems that do not connect
Audited and cleaned
Duplicates, gaps and errors found and fixed
Structured and governed
Owned, documented and access-controlled
AI-ready
Fit to train, fine-tune or feed AI systems safely
Market context
You are not unusual. 43% of Chief Data Officers name data quality as their most significant ongoing priority (Informatica CDO Insights, 2025). 42% of UK companies abandoned most AI initiatives in 2025, up from 17% the year before (S&P Global, 2025). The common factor in those failed projects is not bad technology. It is bad data.
Ready to talk?
30-minute call. No obligation.
Your data is in spreadsheets, in people's heads, in systems that do not talk to each other. You have been told you need 'AI-ready data' but nobody has explained what that actually means. You tried an AI tool and the outputs were wrong, and someone said it is because your data is not good enough.
AI-ready data meets four requirements defined by the Department for Science, Innovation and Technology (DSIT, January 2026): technical optimisation, data and metadata quality, organisational and infrastructure context, and legal, security, and ethical compliance. Basic deduplication and formatting address only part of one requirement. Most organisations need work across all four.
That DSIT definition matters because it means there is no single checklist that makes data 'ready'. Readiness depends on what you want the AI to do. A charity using AI to segment donors needs different data standards than an SME using AI to forecast inventory. A council consolidating service records for predictive analytics has different compliance requirements than a manufacturer cleaning product data.
This is why we start every engagement by understanding your intended AI use cases. Your data gets prepared for specific purposes, not to an abstract standard.
A typical data audit runs two to four weeks depending on how many systems hold your data and how complex the data relationships are. You receive a written report covering what data exists across your organisation, where it lives, what state it is in, and what needs to happen before AI tools can use it reliably.
The full engagement runs across four phases. Each produces a defined output so you always know what you are paying for and what you are getting.
We map every data source across your organisation. Spreadsheets, CRM records, databases, email inboxes, paper files, and the knowledge that lives in people's heads. We identify duplicates, gaps, inconsistencies, and format problems. We assess your current data maturity against what your intended AI applications require.
Output: Written audit report with findings, risk areas, and prioritised recommendations.
We clean, deduplicate, restructure, and standardise your data based on the audit findings. This is hands-on work with your actual datasets. We do the work, not hand you a list of things to fix later.
Output: Cleaned and structured datasets ready for governance and AI preparation.
We put documented ownership, access controls, maintenance responsibilities, and retention schedules around your data. We check that data handling meets UK GDPR (the UK's data protection regulation) requirements and document everything so your team can maintain standards after the engagement ends.
Output: Data governance policy, ownership map, and compliance documentation.
We prepare specific datasets for your target AI applications. This includes formatting, validation, and contextual enrichment where needed. We confirm that the data meets the requirements for your intended use cases and document any remaining gaps.
Output: AI-ready datasets with documentation, plus a written action plan for anything outstanding.
We work with your existing systems. The service is designed for organisations running a mix of spreadsheets, CRMs, legacy databases, and manual records. 46% of UK SME employers use spreadsheets for data management and 57% still use paper-based records (Longitudinal Small Business Survey, 2024). 57% of charities manually key data between systems (Charity Digital Skills Report, 2024). We start from where you are.
If the audit reveals that a system change would genuinely improve your position, we will say so. We do not sell software and we do not take commissions from vendors. Our recommendation is always based on what the data shows, not what generates the largest follow-on project.
81% of UK organisations report being held back by data distributed across multiple systems (Experian, 2024). Scattered data is the norm, not the exception. Phase 1 of the engagement maps every source, including the ones your team has forgotten about. We then build a consolidated picture of what you have before touching anything.
The consolidation work happens in Phase 2. We bring data together, resolve conflicts between sources, standardise formats, and create a single documented view. If your donor records in the CRM say one thing and your finance spreadsheet says another, we find the correct version and reconcile.
For charities managing beneficiary data across grant applications, CRM records, and programme databases, this consolidation often reveals duplication that inflates reported numbers. For SMEs, it surfaces customer records with outdated contact details, incomplete purchase histories, or conflicting account information. For public sector organisations, it identifies the silos that prevent joined-up service delivery.
Data consolidation for AI use is permitted under UK GDPR when specific conditions are met. Those conditions include identifying a lawful basis for each processing operation, conducting a compatibility assessment when repurposing existing data, completing a Data Protection Impact Assessment (a DPIA, the formal risk assessment the ICO requires for high-risk processing), and maintaining transparency with the individuals whose data you hold.
The ICO's (Information Commissioner's Office) published position is that data protection law supports responsible AI development. The Information Commissioner's foreword to the ICO's AI guidance states that data minimisation and AI need not be in conflict. Insightful AI builds these requirements into Phase 3 of every engagement. We document the lawful basis, conduct the DPIA where required, and produce the compliance records your Data Protection Officer or solicitor needs to review.
49% of UK organisations that have not adopted AI cite data privacy concerns as a barrier (YouGov, 2025). That concern is valid, but the answer is documented governance, not avoidance. Our approach to data handling is set out on our security and data protection page.
Insightful AI's data preparation service starts from £3,000 ex VAT as a fixed-price engagement. The scope is defined during a discovery workshop where we assess your data sources, complexity, and target AI use cases. You receive a fixed quote before any work begins.
For context: data preparation accounts for 15 to 25% of total AI project budgets. The question is whether you spend that upfront, where it prevents failure, or discover the cost later when AI tools produce unreliable results. Organisations spend between 10 and 30% of turnover handling data quality issues (UK Government Data Quality Hub, citing DAMA).
A £3,000 data audit can prevent a £30,000 AI implementation from failing. That makes it one of the highest-return investments in your AI journey.
| What you need | What we do | What you get |
|---|---|---|
| Understand what data you have | Data maturity assessment | Written audit report with findings and priorities |
| Fix duplicates, gaps, and inconsistencies | Data quality improvement | Cleaned, structured, deduplicated datasets |
| Know who owns what and how it is maintained | Data governance implementation | Documented governance policy and ownership map |
| Prepare specific datasets for AI use | AI readiness preparation | Datasets formatted and validated for your target AI applications |
| Stay compliant with UK GDPR | Compliance and privacy checks | Data handling documentation meeting regulatory requirements |
The audit phase requires time from your team to explain where data lives and how it is used. This usually means two to three sessions of an hour each during the first two weeks. Beyond that, the hands-on data work is done by our team. We access your systems under agreed protocols and work to a schedule that fits your operations.
For charities running on small teams with no spare capacity, we structure the engagement to minimise disruption. We work with North West organisations where the CEO, the fundraising lead, and the finance officer are all the same person. The process is designed for that reality, not for organisations with dedicated data teams.
For public sector organisations working within procurement and information governance requirements, we provide the documentation your IG team needs to approve access before the engagement begins.
The Phase 4 output includes a written action plan and governance documentation designed so your team can maintain data quality independently. Not every organisation needs ongoing support.
Data does degrade over time. Staff turnover, system changes, and roughly 11% of the UK population moving home each year all erode data quality. If you want periodic reviews to check that standards are holding, we can arrange that. But the engagement is designed to leave you self-sufficient, not dependent.
Where the audit reveals broader organisational issues, an AI strategy engagement is the natural next step. Where it surfaces the need for AI policies or risk documentation, AI ethics and governance runs alongside or ahead of data preparation. And where your data is clean and ready, AI at Work or AI software development is where that preparation pays off.
Myth
'AI will sort out our data problems.'
What actually happens
AI does not fix data problems. It exposes them. If your records are duplicated, your fields inconsistent, or your information locked in someone's inbox, AI tools will amplify those problems. You will get confident-sounding outputs built on unreliable foundations. 80%+ of AI projects fail, and poor data quality is the primary driver (RAND Corporation).
Myth
'Clean data and AI-ready data are the same thing.'
What actually happens
Cleaning removes duplicates and fixes formatting. AI readiness goes further. It requires semantic structuring (so the AI understands what the data means), contextual enrichment (so the AI has enough background to produce useful outputs), bias checking (so outputs are not skewed by unrepresentative training data), and compliance documentation (so use is lawful). DSIT's January 2026 guidance describes four distinct pillars. Stopping at 'clean' means your AI underperforms or produces skewed results.
Myth
'Data preparation is a one-off project.'
What actually happens
Data degrades continuously. A governance policy without maintenance is a document that goes stale within months. Phase 3 of our engagement builds the maintenance processes, not just the initial documentation.
Myth
'Only large organisations need data governance.'
What actually happens
76% of charities lack any formal data strategy (Charity Digital Skills Report, 2024). 43% of public sector organisations lack one entirely (HPE). Size does not determine need. The ICO's expectations apply to every organisation that processes personal data, regardless of headcount.
Myth
'Our spreadsheets work fine.'
What actually happens
They might work for now. They do not scale, they do not enforce consistency, and they create single points of failure. When a team member who built the master spreadsheet leaves, the logic goes with them. 40% of UK SMEs do not have a CRM at all (Sagacity Solutions). The spreadsheet is usually the symptom, not the cause.
Myth
'We need perfect data before we can start using AI.'
What actually happens
You do not. Data readiness is use-case dependent. Different AI applications require different data standards. DSIT (January 2026) states that AI readiness cannot be reduced to a single checklist. Waiting for perfection delays adoption indefinitely. A targeted assessment identifies what needs fixing for your specific goals and leaves the rest alone. Perfectionism is as damaging as neglect.
75% of councils report poor data quality as a challenge to a great or moderate extent (Local Government Association, 2024). Zero said 'not at all'. Your organisation is dealing with a problem that is genuinely widespread. The difference is whether you address it before AI adoption or discover it after.
See how we deliver all our services: How we work | All AI services
Whether you're just getting started or looking to scale, we'll help you find the right path, responsibly.