Gojiberry AI features: what each part does in the workflow

A feature map organized around the work

Gojiberry AI features span discovery, qualification and outreach. Product terms such as agent, source and campaign can sound interchangeable, but they represent different stages. Discovery supplies candidates; qualification decides whether they fit; outreach determines what happens next. Keeping those stages separate makes it easier to diagnose a problem without rebuilding everything.

The following map describes publicly documented capabilities checked on November 9, 2026. It is not a hands-on interface test or a statement that every feature is included in every plan. Where your buying decision depends on a particular setting, ask for a demonstration using the configuration you intend to purchase.

Signals find activity; lookalikes find similarity

Signals are a sourcing approach based on observed activity or business events. Lookalikes start from a reference profile or company and seek similar prospects. The official lookalike guide describes matching attributes such as role and firmographics. These approaches answer different questions: what happened recently, and who resembles an account or person we already understand?

That difference changes how you should review a lead. For a signal lead, ask whether the event is relevant to your offer. For a lookalike, ask whether the shared attributes explain purchasing behavior or merely resemble it. Two companies can have similar headcount and titles while operating with very different budgets, processes and needs.

Choose your reference thoughtfully. A famous customer with an unusual implementation may be a poor starting point for a repeatable audience. A less recognizable account that bought for the same reason as most successful customers may provide a better hypothesis. Keep notes about why the reference is commercially representative.

The low-results help article also documents manual routes such as CSV and existing lists, and notes that Smart Lead Finder may broaden discovery. If you are evaluating signal precision, check whether additional discovery is contributing prospects. Otherwise you could attribute good or weak results to the wrong source.

ICP filtering and scoring turn candidates into a queue

An ideal customer profile is the specification against which a candidate is assessed. In a useful specification, role, company characteristics and exclusions reflect buying reality. A broad label such as technology companies is rarely enough. A narrower operational definition gives the system and the reviewer something observable to assess.

The product’s glossary describes ICP criteria including titles, industries, company size, locations and keywords. The results guide documents stricter and broader filtering modes. This flexibility creates a tradeoff: increasing precision can reduce volume, while broader settings may increase review effort. Neither direction is inherently better without reference to your team’s capacity and economics.

Treat scores as prioritization aids. They are not proof that someone has budget, decision authority or a current project. During a pilot, compare a sample of highly scored and lower-scored leads using your own acceptance criteria. A score that correlates with your definition of a good conversation is more valuable than one that merely looks confident.

Mandatory keywords deserve particular attention. A buyer may have the correct responsibilities without using the exact phrase you expected in a profile. Inspect rejected examples before adding more restrictions, and change one criterion at a time so the effect remains understandable.

AI messages, prompted templates and fixed copy

The documentation describes AI-generated messages using prospect information, activity context and your workspace’s company information. It also distinguishes AI templates, which provide guidance and examples, from fixed messages, which preserve exact text. The right choice depends on how much variation you need and how much factual control the message requires.

For AI-assisted copy, your workspace description becomes a reusable brief. Explain the buyer’s problem, the outcome you can support, the offer and the evidence you are allowed to mention. Remove unapproved claims before they become repeated inputs. A polished tone cannot compensate for an inaccurate proposition.

A template can constrain the structure without forcing a fabricated personal observation. For example, ask for one relevant business hypothesis, one sentence describing the offer and one low-pressure question. Specify that uncertain claims must be omitted. Then review outputs across different roles and company types to see whether the instruction works outside a single example.

Fixed copy is useful when a statement must remain exact, but it still needs a relevant audience. Personalization should add meaningful context rather than decorative references to someone’s job title. Compare how much editing each approach requires before concluding that automatic variation saves time.

Sequences and Review Mode govern execution

Gojiberry documents LinkedIn and email outreach steps, with channel actions dependent on connection state and sender settings. The scheduling guide describes outreach launching twice daily within configured windows. A continuously running source does not imply that every message sends immediately after discovery.

Review Mode introduces a decision point before outreach. For a new audience, use it to validate both the prospect and the proposed communication. Ask during setup whether later sequence steps proceed automatically after approval, and learn how to remove a contact if its circumstances change.

A practical sequence respects dependencies. A follow-up should make sense after the preceding action, and a channel switch should not create a confusing duplicate conversation. Test representative cases: a person who accepts, one who does not, one who replies early and one who should be excluded. These scenarios reveal more than simply counting the available action blocks.

Account-level quotas also matter when several campaigns share a sender. Buying another campaign or source does not automatically mean the same LinkedIn account can send more. Ask which resources are shared before trying to solve a queue problem with additional subscriptions.

Inbox, reporting and integrations complete the handoff

The official product overview lists CRM connections including HubSpot, Pipedrive and Attio, plus API and MCP access for compatible tools. An integration logo is a starting point for evaluation. The important questions concern which objects move, in which direction, on what trigger, and with what access permissions.

Define the expected handoff in plain language. For example, an accepted prospect should have an owner, an account association, a useful source note and a next action. Ask the vendor to show how that outcome is achieved in your CRM. A connection that creates contacts may still require additional work to meet your sales process.

The Inbox helps consolidate conversations, while reporting can help compare sources and campaign results. Use both to follow a prospect through the funnel rather than treating sends as success. A source with fewer contacts may be the better investment if its conversations consistently meet your acceptance criteria.

For API or MCP use, begin with a narrowly scoped task such as inspecting lead context. Confirm authentication, permissions and whether a tool can trigger outbound actions before connecting it to a broader assistant workflow. The value is bringing useful context into the tools your team already uses; the setup still needs an accountable owner.

Frequently asked questions