How to implement enterprise search: three phases to a search people actually use

Enterprise search is software that lets employees search all of a company's internal systems at once, from emails and files to databases, intranets and collaboration tools, through a single search bar. It covers structured and unstructured data and differs from site search on a public website or the intranet search built into a single tool. Unlike traditional search engines, it respects existing access rights, so everyone only finds the relevant documents they are allowed to see. For a full explanation of how it works, read What is enterprise search?.

Rolling out a unified search across every system sounds like a major IT project. In practice, the technology is usually the smaller part. What matters is whether the right data sources are connected, whether permissions are clean and whether employees actually adopt the new search in their daily work.

This guide shows how to implement enterprise search: which resources you need, how the three phases of an enterprise search implementation work, how to onboard users and how to measure the benefits. We describe the phases and timings using amberSearch as the example, because we cannot speak for other vendors' projects. The principles apply to any modern enterprise search solution.

Key takeaways

• Three phases: kick-off and technical set-up (1 to 2 weeks), indexing and key user workshop (5 to 21 days), rollout and onboarding (2 to 4 weeks).
• Little technology, lots of organisation: Connecting a cloud application often takes only one to two hours of technical work. Internal approvals, permissions and supporting users take more time.
• Two roles are essential: IT handles set-up and integration, a project lead takes care of communication and change management.
• Start small: Begin with a few departments of 50 to 100 people, collect feedback, then roll out.
• Measure the benefits: Record search time before the start, then track usage and searches without results after go-live.

Before you start: what you need

With standard software, implementing enterprise search is no longer a huge project. You still need to plan a few resources. They fall into three groups: systems, people and budget.

Systems, hosting and data

The first question is hosting. Three options are common, and not every vendor offers all of them:

• Cloud hosting: The simplest option and the lightest on resources. If you already work in the cloud with Microsoft 365 and other applications, it is usually the best fit. No servers need to be purchased. Depending on the vendor, you pay extra for cloud resources or the number of documents; with amberSearch this is included in the user price.
• Private cloud: Larger companies run dedicated instances on AWS, Azure and similar cloud services. Enterprise search can run there too. No new hardware is needed, but installation requires closer coordination between vendor and IT, because individual settings have to be taken into account.
• On-premises: Some organisations, for example in critical infrastructure, must keep their enterprise data entirely on their own infrastructure. In that case the company pays for the servers, their operation and the know-how to run them.

For private cloud and on-premises, server sizing depends mainly on the volume of data and the number of users.

The second question is make or buy. Open source modules exist as a starting point, but a home-grown search system has to be developed and maintained for good: interfaces change, new capabilities such as AI-powered search and natural language processing keep arriving, and users now expect a search experience as good as web search. Most companies therefore choose ready-made enterprise search software. Many enterprise search tools now combine keyword matching with machine learning and retrieval augmented generation (RAG) to return relevant results as answers. Look for a solution with an open interface so you can connect your own systems if needed. Enterprise search solutions differ mainly in connectors, hosting, access control and AI features, and in whether they use federated search across live systems or build a central index. Our checklist for choosing an enterprise search lists the criteria that matter, and our guide to enterprise search solutions gives an overview of vendors.

The third question concerns your data sources: before the project starts, list the systems where knowledge lives (for example SharePoint, Teams, Outlook, network drives, CRM or ERP) and decide which sources are part of the first release. To begin with, the systems that generate the most questions are enough. Also check which content is outdated or should not be searchable at all.

Team and roles

On your side you need two core resources, plus several groups that play a part in the project:

• IT administrators are responsible for the technical set-up and integration with existing systems.
• The project lead handles communication and change management and is the first point of contact for questions at the start. Even easy-to-use software needs someone who owns the rollout.
• Multipliers are managers and digitally minded colleagues who act as opinion leaders and enable their teams.
• Power users use the search intensively, soon know every feature and become internal experts.
• Regular users use the search to a normal extent. The whole project is for them.

Depending on your organisation, data protection and employee representatives should be involved early. If personal data is processed by a vendor, you will need a data processing agreement under Article 28 GDPR. Only when IT and the project lead work together does a technical installation turn into a search people use.

Budget and business case

The budget depends mainly on the hosting option and the size of the company. Plan for three cost blocks:

• External set-up costs: what the vendor needs for the installation.
• Internal set-up costs: how much time IT, the project lead and business units spend on coordination and approvals.
• Ongoing costs: licences or fees, depending on the model based on users, data volume or number of connected systems, plus maintenance if you host it yourself.

The total cannot be stated in general terms. Our article on calculating a business case for enterprise search describes a method, and the ROI calculator gives you a first estimate.

The three phases of enterprise search implementation

Using amberSearch as a fully managed service as the example (the process changes slightly with other hosting options), the implementation has three phases. The timings are based on experience and cover purely technical effort. Only your company can judge which internal processes and approvals come on top.

Phase 1: Kick-off and technical set-up (1 to 2 weeks)

Goal of this phase: the systems are connected and access rights are synchronised. After a short kick-off, IT connects the data sources to the search:

• Cloud applications are connected via OAuth or a simple app registration, roughly 1 to 2 hours of technical work. Step-by-step guides for each system are in the help centre under data source set-up for administrators.
• Locally installed applications are connected via a site-to-site IPsec VPN. The search also needs a service account with read rights. Again around 1 to 2 hours.
• Open interface: Systems without a standard connector can be integrated by you through an open interface. Most projects do not need this.
• Identity provider: A connection to SSO or Active Directory (or other directory services) ensures that user permissions are carried over. Usually under an hour.
• Domain release: Depending on hosting and the desired domain, an SSL certificate is created, also usually under an hour.

Result: the solution is technically ready but not yet filled with data. R-Biopharm shows how fast this can be: there, the technical connection to the internal systems was completed in coordination with IT within half a day.

Phase 2: Indexing and key user workshop (5 to 21 days)

After set-up, the software builds a searchable index. Depending on the amount of data, this usually takes between 5 and 21 days. Use this time for the key user workshop: a one-hour online meeting with the three or four most important users from different departments. It has three parts:

• Jobs to be done: How does the person work, which systems do they use, what are their daily tasks?
• Pains: What stops them from doing their job well? Where does it get stuck?
• Gains: What value do they expect? When would the rollout count as a success?

The workshop places the search early with future power users and multipliers and shows where the biggest benefit lies. Our article on capturing research time explains how to measure time spent searching at the same time. After that, the first users test the search, usually the project lead, IT administrators and key users, and check that it works as expected.

Phase 3: Rollout and onboarding (2 to 4 weeks)

Now the search is made available to all employees, right where they already work. This has turned out to be one of the most important success factors. The administrator deploys the search centrally, for example as a Teams app, as a widget in SharePoint or your intranet, as a browser extension via group policies or via iFrame in other systems. This takes about an hour and gives everyone a unified search experience in the tools they already use. Users can also add the search to their browser or desktop themselves in five to ten minutes; the options are listed in the help centre under apps and extensions.

After go-live, monitoring begins: search logs, searches without results and response times show where fine-tuning is needed. The next section covers how to win users over in this phase.

Winning users: onboarding that works

An enterprise search is only as good as its adoption. From the user's point of view, the rollout runs in four steps.

1. Key user workshop and indexing: Onboarding for key users already starts in phase 2. They test first and provide the first feedback.

2. Onboarding: Give users access step by step so you can learn from each group and fine-tune. In the onboarding session (around 30 minutes online), you explain to users or multipliers how the search works, they try it out and ask questions. Afterwards everyone receives an onboarding email with the key information. In this phase, employees should

• make the search visible in their daily work and IT set-up,
• get to know the features,
• understand the search logic, refine queries and use filters.

Some people find this easier than others. Short tips with sample searches from their own department help them get started. A regular check-in between the project lead and the vendor's support team in the first weeks helps resolve questions quickly. Material for internal communication is available in the help centre under onboarding.

3. First experiences: A few power users emerge quickly. Their knowledge helps answer other users' questions. Encourage everyone to talk about problems and successes in internal channels.

4. Part of everyday work: A good sign is when the search comes up naturally in meetings. If the solution has not yet been rolled out everywhere, other departments often notice and want to use it too.

Build an internal community

After onboarding, set up a dedicated channel in Teams or another tool for questions about the search. Colleagues answer company-specific questions directly, questions that have already been asked stay easy to find, and successes made visible by the multipliers win over sceptical colleagues too.

Technical pitfalls

Even though modern SaaS solutions make implementation much simpler, three enterprise search challenges come up again and again.

Heterogeneous systems and multiple data sources

Many companies have grown IT architectures: a mix of cloud and on-premises, different API standards, custom developments. The difficulty lies less in connecting multiple systems than in bringing different data models together. That is why a clean inventory of systems before the start, clear prioritisation of data sources and a clearly defined scope for the first release are essential.

Permissions and identity management

An enterprise search takes over existing access rights, including their weaknesses. Inconsistent Active Directory groups, manually maintained SharePoint permissions or orphaned user accounts mean that search results appear unexpectedly or are missing. A short permissions review before indexing is worth it. The goal is not to reorganise your entire identity management, but to have consistent structures that can be synchronised. Our guide on how to optimise SharePoint search shows how to prepare SharePoint in particular.

Data quality and duplicates

Enterprise search makes data quality visible, sometimes brutally so. Duplicates, multiple versions of the same document and inconsistent file names come to light. Managing expectations helps here: enterprise search does not replace document management systems, it exposes inconsistencies. With large amounts of old content, it pays to exclude archives deliberately. Duplicate detection, a ranking boost for current versions and enriched metadata can also help.

Success factors and common mistakes

With almost any software rollout, success depends on making early wins visible. These best practices have proven themselves:

• Consider all user groups: Departments have different needs. Feed the lessons from the first phases back into the project.
• Start small: Begin with a few departments of 50 to 100 people before rolling out to the whole company.
• Market the rollout internally: Announce it and build curiosity so people can prepare for the change.
• Involve local subsidiaries: If you operate internationally, bring in local managers early; they are important multipliers.
• Let users discover use cases: Users should try out for themselves where the search helps them and share their successes openly.

The mistakes that slow a rollout down are just as consistent:

• Too many sources at once: Connecting every data source at launch stretches the project and dilutes the results.
• Permissions not checked: Messy rights only surface when someone sees results they should not see, or important documents are missing.
• Search in the wrong place: A separate web address nobody visits will not be used. The search belongs in Teams, the intranet or the browser.
• No project lead: Without a contact person for questions and communication, the rollout stays an IT project.
• Onboarding as a one-off session: One training is not enough. Users need channels for questions and examples from their own work.
• No baseline: If you do not know how long searching takes today, you cannot prove the benefit later.

Measuring the benefits

Decide before the start how you will recognise success, and record the baseline: how much time do employees currently spend looking for relevant information, and how often do they ask colleagues because they cannot find something? After go-live, the number of active users, searches without results, time to answer and the share of searches with a click on a result show whether the search is being adopted and where it is stuck. Anyone who finds answers faster gains time for their actual work, and that productivity gain is the strongest argument for the project. In customer-facing teams, faster answers can also improve customer satisfaction. Enterprise search is a building block of knowledge management, so its success shows in how well knowledge flows, not in the number of connected systems. Our article on calculating the business case shows how to turn time saved into a business case, and the ROI calculator for AI gives you a first estimate.

Frequently asked questions

How long does an enterprise search implementation take?

It depends on the solution, the hosting and your infrastructure. Using amberSearch as the example: set-up 1 to 2 weeks, indexing 5 to 21 days depending on data volume, rollout and onboarding 2 to 4 weeks. In a cloud environment, implementation typically takes four to six weeks. On-premises projects or complex system landscapes can take months, depending on company size.

Do we need our own IT department to run it?

With a well-designed cloud solution, one IT administrator who handles set-up and administration is usually enough. A self-developed open source solution or your own on-premises hosting, on the other hand, needs a dedicated team.

What resources does enterprise search require?

You need a hosting option (cloud, private cloud or on-premises), access to the data sources, an IT administrator for set-up and integration, a project lead for communication and change management, and a budget for set-up and ongoing licences. With cloud hosting, no hardware is needed.

Will everyone see every document after the rollout?

No. An enterprise search takes over existing permissions via the identity provider. Everyone only sees what they can also access in the source system. That is why it pays to review your permission structure before indexing.

What is the difference between implementation and adoption?

Implementation usually means the technical part: connecting data sources, synchronising rights, building the index. Adoption also covers onboarding, communication and making the search part of everyday work. A project only succeeds when both work.

Conclusion and next step

Implementing enterprise search today is mainly an organisational project with clear phases: clarify resources, connect systems, win users over, measure the benefits. Setting up the enterprise search system is done within a few weeks; success depends on whether employees find their knowledge faster than before.

If you want to see how the three phases would look with your own systems, take a look at amberSearch: it connects your data sources, takes over existing access rights and delivers answers with sources.

Find answers in seconds: try it free for 30 days

Would you rather talk the approach through with us? Then get a demo. Administrators will find all guides in the amber help centre. More articles are on our enterprise search topic page.